GitHub Repository

AI agent governance layer — sign, monitor and control every agent action. EU AI Act · ANSSI · NIST ready.

3 starsPython

PiQrypt – Cryptographic audit trail for AI agents (Ed25519, Dilithium3)

Name: PiQrypt – Cryptographic audit trail for AI agents (Ed25519, Dilithium3)
Availability: InStock
Author: PiQrypt_Fred

by PiQrypt_Fred·Feb 24, 2026·1 point·0 comments

Visit Project View on HN

AI Analysis

●MidBig BrainBold Bet

Ed25519-signed audit trail for AI agents; Dilithium3 optional. Free but early.

Strengths

•Post-quantum Dilithium3 support ahead of threat timeline; Ed25519 + SHA-256 chain is robust design
•Portable format and local-first philosophy (no vendor lock or cloud dependency) appeal to privacy advocates
•Clear tiering with free-forever option removes adoption friction for individuals and small teams

Weaknesses

•Pre-alpha with minimal integration documentation; unclear how it plugs into existing AI frameworks (LangChain, CrewAI)
•Fork detection is conceptually good but design details (choosing canonical branch) not explained; post-quantum overhead unquantified

Post Description

AI agents are making decisions that affect money, infrastructure, and users — but most systems rely on editable logs.

I built PiQrypt to explore whether an AI agent can have a cryptographically verifiable decision history.

Each event is: - Ed25519 signed - SHA-256 hash chained - Fork-detectable - Optionally RFC 3161 timestamped - Optional post-quantum (Dilithium3 / ML-DSA-65)

The goal isn't blockchain or consensus — it's agent identity continuity and tamper-evident history.

MIT licensed, local-first.

https://github.com/piqrypt/piqrypt

I'd appreciate feedback on the design, especially around fork handling and key rotation.