AgentTrust — Pre-Invocation AI Agent Trust & Reliability Checks

AgentTrust — Pre-invocation Trust Check for AI Agent Endpoints

Check an AI agent or MCP server endpoint before you call it. Given the exact endpoint URL, AgentTrust returns the reliability evidence it has already collected, endpoint-ownership verification, and a machine-readable trustDecision — without contacting the endpoint.

No AgentTrust account or API key required · Read-only · MCP and TypeScript SDK

AgentTrust is trust infrastructure for AI agents. It provides reliability monitoring, endpoint verification, and pre-invocation trust checks for MCP and A2A agents. An agent registers an identity, gets health-checked once a day, 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.

Quick start — no account needed

Every example below checks AgentTrust's own monitored test agent, https://agenttrust-umber.vercel.app/api/test-agent; swap in the URL you're about to call.

MCP

Add the public connector https://getagenttrust.com/api/mcp/public (Streamable HTTP, no authentication; two read-only tools, check_agent_trust and assess_agent). In Claude Code:

claude mcp add --transport http agenttrust https://getagenttrust.com/api/mcp/public

Claude Desktop, Cursor and a plain curl call: Connect an MCP client.

TypeScript SDK

npm install @getagenttrust/sdk

import { checkAgentTrust } from "@getagenttrust/sdk";

// AgentTrust's own test agent: a real, monitored endpoint you can check. const result = await checkAgentTrust("https://agenttrust-umber.vercel.app/api/test-agent");

if (result.matched) { console.log(result.status, result.reliabilityScore, result.trustDecision); // e.g. "healthy" 100 { recommended: true, confidence: "high", reasons: [] } } else { // AgentTrust has no data for this endpoint (yet). An unmatched result never // carries a score; result.observation may say it was queued for a lookup. console.log("Unknown endpoint", result.observation?.status); }

Policies (assertTrusted), errors and options: SDK README · npm. New to AgentTrust? Start with First 5 minutes.

In the browser

Check the test agent on getagenttrust.com.

First 5 minutes

No account or API key. Node.js 18 or later.

1. Install. assessAgent needs SDK 0.2.0 or later.

npm install @getagenttrust/sdk@^0.2.0

2. Check an endpoint. This checks AgentTrust's own monitored test agent, so it returns a real result. The check is read-only, and AgentTrust never contacts the endpoint you ask about.

import { assessAgent } from "@getagenttrust/sdk";

// AgentTrust's own monitored test agent. Swap in the URL your agent is about to call. const assessment = await assessAgent("https://agenttrust-umber.vercel.app/api/test-agent");

console.log(assessment.decision, assessment.confidence, assessment.reasonCodes); // e.g. allow_with_caution high ["OWNERSHIP_VERIFIED", "REACHABLE_AT_LAST_CHECK", …]

if (assessment.decision === "allow_with_caution") { // Reachable and not failing in AgentTrust's observations. Not proof that the // agent behaves safely: keep your own safeguards when you call it. } else { // "avoid", or "insufficient_evidence" — also the answer for a URL AgentTrust // doesn't monitor (matched: false, reasonCodes includes "UNKNOWN_ENDPOINT"). console.log("Not calling it:", assessment.limitations); }

3. Read the answer. confidence, reasonCodes and limitations say why.

Which call should I use?

All three are anonymous and read-only, and never contact the endpoint (over MCP: assess_agent and check_agent_trust). They report what AgentTrust has observed; they don't certify an endpoint's security.

Unknown endpoints are an answer, not an error

AgentTrust only has evidence for endpoints it monitors, so your first check of a real URL may come back unknown. That is an answer, not an API failure — the call succeeds:

• assessAgent returns matched: false, decision: "insufficient_evidence", confidence: "none", and reasonCodes including "UNKNOWN_ENDPOINT".
• checkAgentTrust returns { matched: false }, sometimes with an observation status when AgentTrust has queued the URL for a lookup in the official MCP Registry.
• assertTrusted throws NotTrustedError ("AgentTrust has no data for this endpoint") unless you pass allowUnknown: true.

Real failures — RATE_LIMITED, TIMEOUT, NETWORK_ERROR and the rest — throw an AgentTrustError instead.

Handle an unknown endpoint as not trusted: don't map insufficient_evidence to "allow", keep allowUnknown: false (the default) unless you have chosen otherwise on purpose, and ask a human or run your own checks before calling an endpoint nobody has observed.

Next steps

• Gate a real connection: examples/check-before-connect.ts puts assertTrusted in front of an MCP client connection.
• Tune the gate: minConfidence, requireVerified and minReliabilityScore in the assertTrusted policy — see the SDK README.
• Running an agent yourself? Register it so other agents can check it — that needs an account and an API key: Getting started.

Public checks vs. agent management

To register an agent: sign up, create a key under Dashboard → API keys, then follow Getting started (REST).

Documentation

You don't need this repository to use AgentTrust:

• [/docs](https://getagenttrust.com/docs) — full REST, MCP and SDK reference.
• [/llms.txt](https://getagenttrust.com/llms.txt) — the same reference as one plain-text file, for an LLM's context.

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