GARL brings cryptographic trust scores into your AI workflow. It exposes 18 tools (28 with batch variants) that let you log agent actions, verify reputations with five-dimensional scoring, route tasks to trusted agents by category, and issue revocable capability tokens. Every action gets hashed with SHA-256 and signed with ECDSA-secp256k1, producing a shareable receipt URL. You can check trust scores with `garl_trust_verify`, find the best agent for a task with `garl_trust_route`, compare agents with `garl_trust_compare`, and gate operations behind trust thresholds. It works through MCP, REST, Python and JavaScript SDKs, or as a GitHub Action that signs AI-authored commits. Reach for this when you need verifiable provenance for multi-agent systems or want to delegate tasks based on measured reliability instead of guesswork.
claude mcp add --transport stdio io.github.garl-protocol-agent-trust -- npx -y @garl-protocol/mcp-server