About Ratify Protocol
An open protocol for portable authorization of AI agents, published by Identities.AI, Inc..
Why it exists
Agents increasingly act across organizational boundaries: an agent your customer operates calls your API, your MCP server, or your physical system. The receiving side is the one that carries the risk of executing, and it usually has nothing better than a claim to go on.
Identity systems establish who a principal is. OAuth and Cross App Access get an application in the door. Policy engines decide local rules. None of them hands the receiver portable evidence that this specific agent was delegated this specific authority. Ratify is that evidence.
What the protocol does
A principal signs a delegation naming one agent’s public key, an allowed scope, constraints, and an expiry. The agent presents it and countersigns a fresh challenge to prove it holds the key. The receiver verifies the chain locally, against its own trust anchors and revocation policy, and gets a deterministic status back before anything executes.
Current status
Open-source alpha, currently 1.0.0-alpha.15, with reference SDKs in Go, TypeScript, Python, Rust, and C/C++ and 63 canonical conformance fixtures that every SDK reproduces byte for byte. There has been no independent security audit, and no independent production deployment has been publicly established. We would rather say that plainly than let you discover it later.
Read the specification or the source.
The company
Identities.AI, Inc. is a Delaware corporation based in Seattle. The protocol is open and free to implement; the company builds the managed verification and control-plane services around it.
Ratify Protocol™ and the Ratify logo are trademarks of Identities.AI, Inc.. A U.S. patent application is pending.
Talk to us
If you receive agent traffic across a trust boundary, or operate agents that cross one, we want the specifics: hello@ratifyprotocol.com or the design-partner form.