ARC is a proposed authority protocol for delegated AI agent actions, with 14 runnable probes covering delegation, approval, revocation, audit, and retention.

You can run the entire probe suite with a single command:

python3 run_demos.py

At this stage, I’m more interested in rejection than adoption.

If you wouldn’t use ARC, I’d especially like to know why.

Whether it’s a missing use case, the wrong abstraction, an unnecessary protocol, or a flaw in the trust model, every criticism is valuable evidence for improving the protocol.