mcp-audit is itself a security tool. If you find a vulnerability in mcp-audit — for example, a way to crash the scanner with a malicious server response, or an injection that escapes the scanner's report — please report it privately first.

Email: ceo@exsporta.com GitHub: open a private security advisory.

Please include:

• mcp-audit --version
• The MCP server (or a minimal reproduction)
• Expected vs. actual behavior
• Any proof-of-concept

I will acknowledge within 72 hours and aim to ship a fix within 14 days for high/critical issues.

In scope:

• Crashes, hangs, or resource exhaustion in the scanner caused by hostile server responses.
• Report-renderer issues that could lead to injection downstream (e.g. a malicious tool name escaping into a CI log in a way that confuses the surrounding system).
• Privilege escalation when running mcp-audit against a hostile server.

Out of scope:

• Findings produced by the rules themselves — those are the output of the tool. To dispute a rule's accuracy, open a regular GitHub issue.
• Vulnerabilities in the upstream MCP Python SDK — please report those to modelcontextprotocol/python-sdk.

The latest minor release is supported. Older versions receive security backports on a best-effort basis.