Research / Identity and access
OpenAPL
Access policy language for humans and AI agents.
Version 0.1.0Released 2026-11-19
The problem
A support agent hands a refund to a refund agent; the gateway sees only the last hop, so nobody can say which human it's for or what a hijacked agent could reach.
What's new
Evaluates the whole delegation chain, from the human through every agent, and returns ALLOW, DENY or REQUIRE_APPROVAL with the entitlements and control references behind it. No LLM sits in the decision path.
My role
Sole author: designed, built and maintain it.
