MCP config and tool manifest review
Find risky MCP commands, scopes, secrets, and poisoned tool descriptions before an agent uses them.
Paste a client config, server manifest, or tool schema. The audit runs as static analysis only: no command execution, no MCP connection, and redacted evidence in the report.
--
score
- Checks
- SSRF, token passthrough, OAuth, command execution
- Output
- Prioritized findings, fixes, copyable report
- Mode
- Paste-only static review
Findings
Prioritized review queue
Audit output appears here with evidence and remediation steps.
What this checks
- Shell/interpreter execution paths, package runners, and unpinned dependencies.
- Secret-looking env vars, pasted tokens, broad filesystem paths, and runtime privilege flags.
- SSRF-prone metadata/internal URLs, wildcard redirects, broad OAuth scopes, and token passthrough language.
- Tool-description poisoning and risky schemas for path, URL, SQL, regex, or command parameters.