Before one agent delegates work to another, someone has to answer a simple question: can I trust this thing?
Enter an agent domain. We check /.well-known/agent.json, mcp.json, acp.json, llms.txt, DNS, and TLS.
Agents are about to move money, sign contracts, and book resources on behalf of millions of businesses. The question isn't whether autonomous commerce happens — it's whether there's a trust layer underneath it when it does.
We probe every public agent discovery endpoint — /.well-known/agent.json, mcp.json, acp.json, /llms.txt, DNS _agent records, and TLS — then classify what we find as verified agent, partial agent, or not detected. Domains without a valid protocol card get their Honest score capped. No signup. No integration. If your agent publishes a card, it already has a score.
No ML. No sentiment. All empirically observable.
reputation.md exposes an MCP tool endpoint at /api/mcp that any MCP-compatible agent (Claude, etc.) can call directly. Two tools: get_agent_score to look up a cached score, and scan_agent to trigger a fresh scan. No API key required.
Prefer REST? POST /api/scan with {"domain":"your-agent.ai"} returns the full score breakdown as JSON. GET /api/agent/:domain returns cached results instantly.
Discover our endpoints: /.well-known/agent.json · /.well-known/mcp.json · /llms.txt
When an agent can verify trust in 50 milliseconds, it can delegate to agents it's never met. New suppliers, new geographies, new capabilities — all accessible because someone kept score.
Credit scores didn't just measure trustworthiness. They made it possible for strangers to do business. This does the same thing at machine speed, for every permutation of agent-to-agent commerce.
The score is the primitive. Commerce is the emergent behavior.