MCP Shield

Documentation

Learn how to use MCP Shield, understand scan results, and integrate the scanner into your development workflow.

Quick start

  1. Open Scan Your MCP.
  2. Upload an MCP tool/server JSON file, or choose an Acme DevTools sample.
  3. Review the AI Trust Decision, risk score, findings, and recommendations.
  4. Ask the AI Security Copilot to explain any finding in plain language.
  5. Optionally generate an advisory hardened rewrite — the original config is never modified.

What a scan returns

  • Deterministic risk score (0–100) and security grade
  • Evidence-backed findings with OWASP MCP mappings
  • Permission analysis, combination risks, and attack-path simulation
  • AI trust decision and deployment recommendation
  • Business impact categories and prioritized recommendations

API surface

  • POST /api/scan — run the full scan + copilot insights pipeline
  • POST /api/copilot/chat — ask questions grounded in a ScanResult
  • POST /api/copilot/harden — advisory hardened rewrite + re-scan
  • GET /api/samples/[name] — load a sample MCP config
Back to Home