Show the agents and developers who connect to your MCP server that its advertised tools are free of prompt-injection and tool-poisoning. The badge re-scans every six hours and links to the full audit.
A real public MCP server, scanned right now (static metadata only — we never call its tools):
https://mcp.deepwiki.com/mcp
clear — no injection/poisoning patterns in any advertised tool. review — suspicious metadata worth a human look. block — do not admit this server until resolved.
Want the itemized findings — which tool, the injected text, the OWASP mapping (LLM01/LLM08) and remediation? That's the full audit: GET /scan/mcp?url=… ($0.05 USDC via x402), or zero-egress POST /scan/mcp with your tools/list for private servers.
Gate every build on the verdict. A subscription gives you an API key with 100 scans / month across any Grey Ridge endpoint — call GET /scan/mcp?url=… (or POST your tools/list for private servers) from your pipeline and fail the build on block. Pay by card — no wallet, no crypto.
Prefer per-call with no account? Agents pay inline over x402 at $0.05/scan — the free badge above keeps re-scanning every six hours regardless.