As AI agents begin executing real-world actions (calling APIs, writing to databases, triggering workflows), there's currently no native identity or authorization layer for them.
AIP is an open protocol that gives every agent a cryptographic identity and requires actions to be signed and verified before execution.
The goal is simple: if an agent wants to act, the system should be able to verify who it is and what it's allowed to do — in milliseconds.
The repo includes a Python SDK, verification pipeline, and a decorator that can protect agent actions with a single line.
Curious to hear what the HN community thinks about identity and trust layers for autonomous systems.