HN Showcase
Back to browse
GitHub Repository

Agentic Power of Attorney (APOA) | Authorization infrastructure for AI agents

11 starsTypeScript

Agentic Power of Attorney (APOA) – An open standard for AI agent auth

by juanfiguera·Feb 26, 2026·1 point·0 comments
Visit ProjectView on HN

AI Analysis

●●●BangerBig BrainZero to One

Formal auth framework for agentic delegation; fills a gap before agents negotiate house deals.

Strengths
  • •Zero-capability delegation (ZCAP-LD) is genuinely novel in agent context—audit trail + revocation solves real liability
  • •Builds on proven foundations (OAuth 2.1, JWT) instead of inventing crypto—pragmatic standards work
  • •Real motivating incident (agent sent email to wrong recipient) shows this solves today's problem, not theoretical one
Weaknesses
  • •Working draft with no reference implementations shown—unclear if integrations with major services exist yet
  • •Adoption friction: requires service providers to implement; early-stage without buy-in from Anthropic, OpenAI, or platforms
Category
Security
Target Audience

AI infrastructure builders, auth standards bodies, enterprise agent deployers

Similar To

OAuth 2.1 · ZCAP-LD · WebAuthn

Post Description

"Buy me a house. Budget $475K. Good school district. Handle it." That's where AI agents are headed. One already bought someone a car last month, negotiating $4,200 off across dealerships via email. Its entire authorization framework was a prompt: "ask me before doing anything consequential." It also sent a confidential email to the wrong person.

APOA is the missing infrastructure: an open standard for formally delegating bounded authority to AI agents. Scoped permissions, audit trails, instant revocation, credential isolation. Builds on OAuth 2.1, JWT, ZCAP-LD.

Working draft, Apache 2.0. Looking for feedback from anyone building agent infrastructure or working on auth standards. Please poke holes at it!

Similar Projects