We wrote the standard. Vortalis is the reference implementation.

The Agent Governance and Action Protocol (AGAP™) is an open standard that defines the runtime governance requirements for AI agent systems. It specifies how agents must authenticate, how their actions must be authorised, how sensitive data must be protected, and how every decision must be recorded.

AGAP™ was created by MTE Software Ltd and published under CC BY 4.0. The full specification, including requirement definitions and conformance criteria, is available at agap-protocol.org.

Conformance is structured in three tiers (Core, Extended, and Full), allowing organisations to adopt governance incrementally and verify each level independently.

Conformance per control is published, with the difference between letter-code-bound and capability-bound rows on the page rather than buried in the test suite.

Every requirement is recorded in one of three categories. A requirement is letter-code-bound when a named test file under tests/conformance/ binds it, so a regulator can read the requirement, the assertion, and the implementation the assertion exercises together. It is capability-bound when the implementation exists and is exercised by the broader test suite but no dedicated letter-coded test has been bound yet. The remainder is not represented, where neither a binding nor a clear implementation pointer exists today. The binding programme is in progress, and capability-bound rows convert to letter-code-bound as named test files land.

A capability-bound row is not marketed as if it were letter-code-bound. The categorisation lives in public so the difference is on the page rather than buried in the test suite. Every Vortalis release is exercised against the AGAP™ Conformance Test Suite in CI, alongside 1000+ platform-wide tests.