AgentTrust — Pre-Invocation AI Agent Trust & Reliability Checks MCP server

AgentTrust

AgentTrust is trust infrastructure for AI agents. It provides reliability, reputation, endpoint verification, and pre-invocation trust checks for MCP and A2A agents. An agent registers an identity, gets continuously health-monitored, optionally proves ownership of its endpoint, and accumulates a deterministic reliability score from that observed history. Any external AI agent or system can look up another agent by its invocation URL and get back a machine-readable trustDecision — a signal derived from AgentTrust's own observed and verified data, for the caller to weigh, never a certification or guarantee of safety — before deciding whether to interact with it.

Official MCP Registry identity: io.github.agenttrust-ai/agenttrust.

Using the API

If you're building an AI agent or system that wants to look up another agent's trust information, you don't need this repository at all — see:

MCP reference, with real request/response examples.

• /docs — full REST and

same reference as a single plain-text file, meant for pasting into an LLM's context or fetching programmatically.

• /llms.txt — the

Anonymous pre-invocation trust check (MCP)

The fastest way to evaluate an agent before invoking it needs no AgentTrust account or API key at all — call the checkagenttrust MCP tool at /api/mcp:

check_agent_trust({ "endpointUrl": "https://the-agent-you-are-about-to-call.example.com/invoke" })

invoke or otherwise contact the target endpoint during the lookup.

• Read-only.
• No AgentTrust account or API key required.
• Checks only AgentTrust's already-stored observations — it does not

confidence, reasons) for the caller to evaluate.

• Returns a machine-readable trustDecision (recommended,

Try it with a known endpoint

To see the lookup flow work end-to-end without registering anything yourself, call checkagenttrust against an endpoint AgentTrust already observes:

check_agent_trust({ "endpointUrl": "https://allagents.app/a2a" })

This endpoint is already tracked by AgentTrust, so the call exercises the real lookup path against real observed data. Its health status, reliability score, verification state, and trustDecision can change over time — the point here is to confirm the flow works, not to check that endpoint's current standing.

Authenticated REST/MCP operations

Registering an agent, or reading the fuller per-agent record (agent card, capabilities, health history), requires a human-created API key: sign up, create a key in the dashboard, then GET /api/v1/agents?endpointurl= with Authorization: Bearer — the response includes the same trustDecision. The authenticated MCP tools (listagents, getagent, getagenthealth, sendheartbeat) mirror the REST API exactly.

Developing this project

This is a Next.js (App Router) + Drizzle + Supabase app, deployed on Vercel.

npm install
npm run dev

Open http://localhost:3000. You'll need a .env.local — see .env.example for the required variables (Supabase credentials, database connection strings, and the server-only secrets described inline).

Useful scripts:

npm run test        # vitest — full suite runs against an embedded pglite DB, no live database needed
npm run typecheck    # tsc --noEmit
npm run lint         # eslint
npm run db:generate   # generate a new Drizzle migration from schema changes (offline)
npm run db:migrate    # apply pending migrations to the database in DIRECT_URL