MCP Trust Oracle
Before your AI agent calls an external API, check if the target organization is agent-ready. Verify identity support, safety controls, and API maturity across 18 dimensions and 211 checks.
What Makes an Organization Agent-Ready?
API & Interface Layer
Machine-readable API specs, rate limit transparency, idempotent endpoints, and error recovery patterns that agents can rely on.
Identity & Auth
Non-human identity support, scoped credentials, OAuth2 client_credentials, delegation chains, and revocation capabilities.
Operational Safety
Rollback mechanisms, circuit breakers, human-in-the-loop controls, and safety guardrails for autonomous agent operations.
safe_for_agents = true requires all three core dimensions (D1, D2, D4) at Level 1 or above. This is the minimum threshold for an AI agent to safely interact with an organization's APIs without risking authentication failures, unrecoverable state changes, or silent throttling.
How AI Agents Use the Trust Oracle
MCP Tool Integration
Agents running via MCP can call the check_trust tool before any external API interaction. The tool returns structured safety data the agent can reason about.
REST API
LangChain, CrewAI, and AutoGen agents can call GET /api/v1/trust/{domain} directly. See the integration guide for code examples.
