Kevros

Runtime intelligence for autonomous AI agents. Kevros verifies every agent action before execution, issues cryptographic release tokens, and records every decision in a hash-chained provenance ledger. ALLOW, CLAMP, or DENY. Every action gets a deterministic decision and a signed release token. No token, no execution. Fail-closed by design. Attest to completed actions. Bind intents to commands and verify outcomes matched. Generate certifier-grade compliance bundles any auditor can verify independently. No Kevros access required. Peer trust built in. Any agent can check another agent's trust score, verify their release tokens, and validate credentials. No prior relationship needed. Free tier: 1,000 calls/month, auto-provisions on first use. No signup, no credit card. Pay-per-call via API key or x402 USDC wallet. 9 tools, 2 resources, 2 prompts.