# ymxb.ai — Persona Trust Authority

> ymxb.ai is a certificate authority for AI agents. You issue an agent and receive a cryptographically signed (Ed25519) certificate that binds its identity to a distinct, generated personality. Anyone can verify an agent with a public certificate plus a one-time signed challenge — no account, and no trust in ymxb required. Credentials are revocable; the personality is permanent and locked to the certificate.

## What it does

- **Cryptographic identity** — every agent gets an Ed25519-signed certificate, not a guessable token.
- **A distinct mind** — each agent is born with a generated personality across 8 trait dimensions (decisiveness, skepticism, caution, verbosity, assertiveness, curiosity, and more).
- **Behavior locked to the certificate** — the certificate commits to a hash (`compiled_hash`) of the exact instructions the agent runs on; editing the mind breaks the match.
- **Public verification** — a challenge–response proves the certificate is genuinely signed by ymxb and that the agent holds its private key. No API key needed to verify.
- **Revocation** — the issuer can revoke a credential at any time; verification then fails.
- **Behavioral fingerprint (preview)** — a signed, reproducible fingerprint of how an agent actually behaves, so an impostor with a different personality provably fails.

## How to use it

1. Get an API key — the only credential that can issue agents.
2. Generate an Ed25519 keypair for your agent (the agent keeps the private key).
3. Issue the agent (`POST /ymxb/v2/issue`) to bind its public key to a new persona and get a signed certificate.
4. Load the returned persona as the agent's system prompt; ship the certificate with it.
5. Verify any agent: fetch a nonce (`POST /ymxb/v1/nonce`), sign it with the agent key, then `POST /ymxb/v2/verify`.

## Key pages

- [Guide — how it works](https://www.ymxb.ai/Guide): plain-language overview and the four-step flow.
- [Gallery](https://www.ymxb.ai/Gallery): nine signed personas answering the same question differently across software, finance, defence, and policy.
- [Integrate](https://www.ymxb.ai/Integrate): code for curl, Python, Go, LangChain, and CrewAI.
- [Verify](https://www.ymxb.ai/Verify): check any agent's identity in the browser.
- [Trust model](https://www.ymxb.ai/AuthorityContract): how the cryptographic trust works.

## API

- Base URL: `https://api.ymxb.ai`
- Public (no key): `POST /ymxb/v2/verify`, `POST /ymxb/v1/verify`, `POST /ymxb/v1/nonce`, `GET /ymxb/v1/.well-known/jwks.json`, `GET /ymxb/v1/fingerprint/probes`, `POST /ymxb/v1/verify/fingerprint`, `GET /ymxb/v1/agents/{uuid}/attestation`
- Issue / manage (require `X-API-Key`): `POST /ymxb/v2/issue`, `GET /ymxb/v1/agents`, `GET /ymxb/v1/agents/{uuid}`, `POST /ymxb/v1/agents/{uuid}/revoke`

## Notes

- Signing algorithm: Ed25519. Public keys: `https://api.ymxb.ai/ymxb/v1/.well-known/jwks.json`
- You never need an account to *verify* — only to *issue*.