About
Why this exists.
Models have become capable planners; reliably acting on those plans against real systems has not kept pace. Most agent platforms expose tools to the model and rely on the model to use them correctly. Execution Protocol takes a different approach: every action passes through a deterministic policy gate, is bounded by an explicit delegation, and produces a signed receipt that any third party can verify offline.
What we believe
- Models generate intentions; protocols commit actions. A policy gate between the model and the systems it touches is what separates an agent that demos well from one that runs in production.
- Determinism is the basis for verification. If the Authorization Boundary is non-deterministic, the actions it allows cannot be audited. Determinism is what makes “what was authorized” a question with one answer.
- Receipts are stronger evidence than logs. A signed receipt is verifiable by any third party, offline, without trusting the issuer. For autonomous action, that distinction matters.
Try it
The in-browser sandbox runs the protocol end-to-end with no signup and no key. The open specification covers every wire-format detail needed to implement against the protocol.