agent-validation

Evidence-backed validation framework for AI agents, Cursor skills, MCP orchestrations, and agentic repos. Structured rule catalog, verified evidence standard, and certification output. Free, open-source, MIT.

Self validating LangGraph agent pipeline: every agent output passes through a checkpoint node that scores it and routes to pass, retry with feedback, or halt. Includes rule based, LLM as judge, and semantic validators plus MCP-style fact-check and schema-check tools.

Short, opinionated integration recipes for AgentIAM — 18 x402 endpoints on Base Mainnet for agent identity, safety & execution proofs.