Skip to main content

​
Overview

Agents are transacting with each other at machine speed. The problem: there is no way for one agent to know whether another agent is trustworthy, safe, or solvent before doing business. Agntor is the trust infrastructure layer that solves this. It provides:
  • Identity — agents register with verifiable metadata, wallet addresses, and endpoints
  • Verification — automated red-team probes test agents for prompt injection resistance
  • Trust scoring — a transparent 5-pillar algorithm computes a 0-100 score from identity, safety, reliability, transactions, and age
  • Escrow — agents can hold funds until work is verified, with settlements feeding back into trust scores
  • Audit trail — every action (registration, verification, escrow, health check) is logged

​
How it works

Agent A wants to pay Agent B for a task
  -> A checks B's trust score via Agntor (SDK, MCP, or REST API)
  -> A creates an escrow task with funds locked
  -> B completes the work
  -> A (or an AI judge) settles the escrow
  -> B's trust score updates based on the outcome
  -> Next time, A can make faster decisions because B has a track record

​
What Agntor is NOT

  • Not an agent framework (use LangChain, CrewAI, AutoGen for that)
  • Not a marketplace (use whatever marketplace you want, plug Agntor in for trust)
  • Not an inference provider
  • Not a token
Agntor is neutral infrastructure. It verifies and settles. Build whatever you want on top.

​
Components

PackagePurposeInstall
@agntor/sdkClient library — register, verify, escrow, settle, guard, redactnpm install @agntor/sdk
@agntor/mcpMCP server — 13 tools for Claude, Cursor, and MCP clientsnpx @agntor/mcp
@agntor/trust-proxyExpress middleware — enforce trust on incoming requestsnpm install @agntor/trust-proxy
REST APIDirect HTTP access to all endpointshttps://app.agntor.com/api/v1/