Your AI agents use MCP, A2A, ANP — each with separate credentials, separate revocation, separate audit logs. AIB Cloud gives you one dashboard to manage them all.
Every AI protocol built its own identity layer. You're managing 4 systems that don't talk to each other.
Four patterns. Pick the one that fits your architecture. No lock-in — switch anytime.
Import AIBClient, create a passport, send requests. Protocol detection, credential injection, and audit logging are automatic. 5 lines of Python.
Run aib serve as a sidecar. Change the URL your agents call — that's it. Zero code changes. The gateway handles detection, credentials, and logging.
Register AIB in your Identity Provider. Agents authenticate via Entra/Okta/Auth0, receive an AIB passport. Capabilities from roles. No new identity system to manage.
Got a protocol we don't support? Drop a Python file in aib/bindings/. Three methods, ~50 lines. Auto-discovered on startup.
Built for production. Every feature has tests. Every attack vector has a mitigation.
One identity, all protocols. RS256 signed. Three tiers: permanent, session, ephemeral. Cascade revocation.
A2A ↔ MCP ↔ DID automatic conversion. Input + output validated against injection attacks.
Auto-detection, credential injection, SSRF defense. Adds < 3ms. Scales to 50K req/s.
Every action logged with cryptographic receipts. Merkle Tree proofs. Real-time monitoring.
Entra, Okta, Auth0, Keycloak. Your agents use your existing IdP. No new identity system.
M-of-N signing policy. Gateway + OIDC + admin. Forging requires M keys.
90-day scheduled. Emergency rotation with instant revocation. Zero downtime.
Crypto-shredding, PII guard, portability export, consent tracking. CNIL-validated approach.
New protocol = 1 file. Auto-discovered. 4 built-in (MCP, A2A, ANP, AG-UI).
The protocol is free forever. The managed service is what you pay for.