CVE-2026-57134

PraisonAI, an open-source multi-agent framework, fails to validate credentials in MCPSecurity.evaluatePolicy() (src/praisonai-ts/src/mcp/security.ts): the configured auth.validate() callback is only invoked for api-key and bearer auth methods, while basic and OAuth policies accept any non-empty Authorization header and immediately return an authenticated result. An unauthenticated attacker on the network can send an arbitrary or malformed Authorization header to an MCP endpoint protected by a basic or OAuth policy and be treated as authenticated. This grants access to MCP tools and resources intended to be protected, with high confidentiality impact and low integrity impact (CVSS 3.1: 8.2). Versions 1.5.1 through 1.7.1 are affected when deployments use the praisonai-ts MCP security module with basic or OAuth policies; the flaw is fixed in version 1.7.2. No public PoC exists and no exploitation has been observed in the wild.

What to do: Upgrade PraisonAI to version 1.7.2 or later, where credential validation is enforced for all auth methods. If immediate upgrade is not possible, switch MCP policies from basic/OAuth to api-key or bearer (which do call auth.validate()), or front MCP endpoints with an authenticating reverse proxy. Audit logs for MCP requests with basic or OAuth policies that were accepted between version 1.5.1 and the upgrade, since any non-empty Authorization header would have been honored.