GitHub Repository

Open protocol for cryptographically verifiable human intent — AI agent authorization, local-first, offline-capable, no central authority.

2 starsRust

Oath – prove a human authorized each action before your AI agent acts

Name: Oath – prove a human authorized each action before your AI agent acts
Availability: InStock
Author: danbitengo

by danbitengo·Mar 3, 2026·1 point·2 comments

AI Analysis

●●●BangerBig BrainZero to One

Cryptographic proof-of-authorization for AI actions; solves real agent oversight problem with clever simplicity.

Strengths

•Signal-on-absence is itself evidence—unattempted actions prove human authorization is required, not just checking a bit.
•Local-first design means no trusted server needed; verification works offline indefinitely via UUID proof IDs.
•Demo shows practical integration: Python one-liner for agent, readable audit trail, structured CLI interface.

Weaknesses

•Requires agents to be built expecting attestation checks—retrofitting into existing systems needs refactoring.
•No mention of recovery if a human loses their keypair; UX for key management not addressed.