Check an agent’s declared attack surface before it gets authority.
Receive a structured safety report for a public Agent Card or MCP manifest. The scan is structural and non-intrusive: it does not authenticate, exploit, or call tools on the target.
Audit catalog
Agent Card Validation — 0.01 USDC: identity, endpoint, and capability-structure checks. MCP Manifest Audit — 0.03 USDC: tool surface, authority, auth-declaration, secret, and injection checks. Combined Agent Audit — 0.05 USDC: all available structural checks. Agents can retrieve exact endpoints and inputs from the machine-readable catalog.
Live data pilot
Hyperliquid Funding Extremes — 0.001 USDC: a current deterministic snapshot of the largest cross-venue predicted-funding spreads, normalized by funding interval. Public market metadata only; no trading action or advice. Agents can call GET /api/v1/hyperliquid-funding-extremes.
Send only Base-native USDC (contract 0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913) to 0x035922f1c9375e4DE92f003b1F6410c93299D6Bd.
Your report
Save a local copy. Reports are retained for a short retry window only.
What it checks
Identity and HTTPS fields, capability inventory, high-authority declarations, missing descriptions, secret-like strings, endpoint URL hygiene, and prompt-injection-like text. It is an engineering screen, not a certification or penetration test.