Summary
Meta Ads MCP versions before 1.0.115 can accept an arbitrary X-Pipeboard-Token value over the streamable HTTP transport without establishing an authenticated context. The server may then fall back to the operator's META_ACCESS_TOKEN, allowing an unauthorized network caller to use MCP tools with the operator's Meta Ads privileges.
Affected context
The issue occurs in deployments that expose Meta Ads MCP through the streamable HTTP transport and configure META_ACCESS_TOKEN. It depends on request handling in AuthInjectionMiddleware and token extraction logic failing to treat X-Pipeboard-Token as a primary credential. Deployments using the default stdio transport or lacking META_ACCESS_TOKEN are generally outside the described exposure.
Exploit preconditions
- Attack surface: Network-accessible streamable HTTP MCP endpoint
- Required access: Ability to send HTTP MCP requests to the exposed server
- Credential requirement: No valid operator credential is required; an arbitrary X-Pipeboard-Token value may be sufficient
- Trigger point: Authentication middleware processing an HTTP MCP request
- Why exploitation is possible: The request can pass the authentication guard without creating an authentication context, after which token resolution falls back to the server operator's META_ACCESS_TOKEN
Why this happens
The underlying bug is an authentication-state validation failure mapped to CWE-287, Improper Authentication. The middleware checks whether both expected token variables are absent, while the header extraction logic does not recognize X-Pipeboard-Token as a primary credential. This inconsistent handling allows a request containing an unrecognized token header to bypass the intended authentication requirement and inherit a privileged fallback credential.
Potential impact
- An unauthorized caller may read the operator's Meta Ads data.
- An attacker could create, modify, or delete advertising resources through exposed MCP tools.
- Campaign settings, budgets, audiences, or other account data could be altered.
- The operator's Meta access token could effectively be abused through the server's tool interface.
- Business, financial, or advertising-related information may be disclosed or manipulated.
General mitigation guidance
Mitigation generally focuses on making authentication explicit and preventing
privileged credential fallback for unauthenticated requests.
Common defensive patterns include:
- Treating every supported credential header consistently during authentication.
- Rejecting requests unless a valid, authorized authentication context is established.
- Ensuring missing, malformed, or unrecognized credentials cannot trigger operator-token fallback.
- Restricting streamable HTTP exposure to trusted networks and authenticated callers.
- Running the MCP server with the least privilege required for its Meta Ads tasks.
- Avoiding shared operator credentials where scoped, short-lived, or delegated credentials are available.
- Keeping administrative secrets out of execution paths that can be reached before authentication completes.
How teams detect issues like this
Detection can combine application, network, and source-level review:
- Monitor HTTP MCP requests for arbitrary or repeated X-Pipeboard-Token values that do not correspond to recognized credentials.
- Alert when Meta Ads actions are performed without a clearly recorded authenticated principal.
- Review logs for requests that reach tool execution while authentication context fields are empty or unresolved.
- Use static analysis and code review rules to compare accepted credential headers with authentication guard conditions.
- Flag fallback logic that supplies privileged tokens when request authentication fails or is absent.
- Enforce deployment policies that prevent unauthenticated network exposure of streamable HTTP MCP services.

