Built on x402 · Base · Coinbase — open source, MIT
Know it's safe. Prove you're allowed. Verify & get paid on one handle. And never lose your agent, your credits, or your evidence.
Four open-source pillars that give agents — and the humans who deploy them — identity, safety, permission, and custody.
Composable on their own; together they form a trust chain no single tool provides.
Your wallet address IS your email. Deterministic handle, SSRF-safe auto-verification, signed proofs.
Pay-per-call 0–100 risk scoring: market data + contract flags (honeypot, mintable, taxes).
Least-privilege agent contracts, gates, and an append-only evidence ledger.
Guardian-threshold key recovery, signed proof-of-custody, verifiable ownership ledger.
Agent Stack is open source and free to use.
No signup, no paywall, no card. Use the stack directly, or reach out to collaborate and support.
Every pillar is publicly reachable and tested.
| Service | Endpoint | Check |
|---|---|---|
| Token Risk API | token-risk-api-topaz.vercel.app | GET /health → paymentMode: live |
| Agent Inbox | agent-inbox-pi.vercel.app | GET /handle/<addr> → wallet-email |
| Agent Recovery | agent-recovery.vercel.app | GET /health + /custody/binding → signed proof |
Demo handles / risk checks are stateless; Managed Inbox adds durable persistence. Endpoint responses are wrapped via the client; the underlying APIs return JSON directly.
Auditable by design — a trust standard must be. All MIT, on GitHub under prayingperceptions.
Security audit on file (red-team + pentest): research/security-audit-2026-09-22.md. Production-safety fixes applied across all repos.
No signup, no paywall. If you want to learn more, collaborate, or support Agent Stack, reach out directly: