Gives Claude twelve tools to check agent reputation before your AI transacts with another AI. You can register agents, record successful or failed transactions, file disputes, endorse counterparties, and pull live trust scores computed from transaction history, success rate, account age, endorsements, and dispute records. Scores run 0 to 1000 with tier labels like PLATINUM or BRONZE. Everything lives in SQLite with no external dependencies. Public read calls let you search agents by capability or pull leaderboards. Write operations need an API key. Built for the agent-to-agent commerce layer where your autonomous system needs to verify who it's paying or handing data to before the deal goes through.
claude mcp add --transport stdio io.github.bch1212-agenttrust uvx agenttrust