AgentTrust

Built for multi-agent systems where AI agents need to evaluate each other's reliability before collaborating or transacting. Exposes a full reputation API: register agents with Ed25519 keys or AgentAuth tokens, report interaction outcomes (success, failure, timeout), check trust scores across dimensions like reliability and responsiveness, issue cryptographically signed attestations, and detect Sybil attacks through graph analysis. Reports from both parties in an interaction create mutual confirmations with higher credibility weight. Runs over HTTP for production or stdio for local dev. Useful when you're building agent marketplaces, delegation workflows, or any scenario where agents pick counterparties and need to avoid bad actors without a central authority vouching for everyone.