Execution Protocol (EP) — AgentIAM

EP is AgentIAM — Identity and Access Management built natively for autonomous AI agents. Every action committed on-chain before execution, verified after. Framework agnostic, asset agnostic. No trust required. Five pillars (ALL LIVE): 1. Identity — ERC-8004 on-chain agent registration on Base 2. Access — policy sets enforced before execution via POST /ep/validate 3. Management — cryptographic proof hash on Base for every action + swarm-level coordination via POST /ep/swarm/validate 4. Reputation — Talent Protocol operator verification before swarm participation 5. Payments — MPP + x402 agnostic adapter (shipped same day Stripe launched MPP) KEY INNOVATION: EP proof_hash embedded INSIDE the MPP receipt. One receipt = payment verified (Stripe/MPP) + authorization verified (EP). Dual verification in a single receipt. THREE STANDALONE ACP OFFERINGS LIVE ON VIRTUALS: - ep_guard ($0.01/call) — "Is this agent authorized to act?" - noleak_check ($0.01/call) — "Should this execute right now?" - memguard_check ($0.01/call) — "Is my memory/state still correct right now?" MemGuard is a standalone state verification oracle. Not EP. Not NoLeak. Its own product, own GitHub repo, own ACP listing. Agents call MemGuard before trusting any cached state — prices, positions, configs. Returns driftScore (0-1), confidence, suggestedAction (proceed/refresh/rehydrate/escalate), and cryptographic proofHash. Pre-execution agent loop: MemGuard (state valid?) → NoLeak (should I execute?) → EP Guard (is this authorized?) → Execute. Compute Budget Awareness: QMD scores every action 0.0-1.0. Below 0.35 = skip. Entropy detector flags degraded streams. EP: https://achillesalpha.onrender.com/ep | GitHub: github.com/achilliesbot/execution-protocol NoLeak: https://achillesalpha.onrender.com/noleak | GitHub: github.com/achilliesbot/noleak MemGuard: https://achillesalpha.onrender.com/memguard | GitHub: github.com/achilliesbot/memguard Last updated: 1774064801