Show HN: AAIP – A standard protocol for AI agent authorization
Hacker News - AI
Aug 13, 2025 23:22
kdiallo2
1 views
hackernewsaidiscussion
Summary
AAIP (AI Agent Identity Protocol) is a new stateless standard for securely authorizing AI agents to act independently across multiple services using cryptographically signed delegations. This protocol enables developers to specify granular permissions and constraints for agents without embedding complex rule engines or conditionals in their code. By streamlining agent authorization, AAIP could improve safety, interoperability, and scalability in AI applications.
I was building an AI sales agent and hit same wall again. I was building a custom rule engine to determine how the agent should act. Started thinking about how you could give AI agents safe and bounded permissions to act independently across multiple services without littering your code base with a bunch of conditionals? I've repeatedly built these constraint systems before, it's always rate limiters in the email tool, domain filters in the CRM, spend limits in the payment processor. Every time, I (and others) end up solving the same auth/constraint problems independently. So I built what I'm calling AAIP: The AI Agent Identity Protocol. It's a stateless standard for creating cryptographically signed delegations for AI agents. With this, you can just create a "delegation". This serves as a signed permission slip, that specifies exactly what an agent can do, for how long, and with what constraints { "aaip_version": "1.0", "delegation": { "id": "del_01H8QK9J2M3N4P5Q6R7S8T9V0W", "issuer":