Act on this, mostly signal

Mnestiq records tamper-evident logs of what an AI agent did

What to do

Try the Mnestiq SDK on Anthropic or OpenAI clients if you need signed audit trails.

Why

The logs are Ed25519-signed and hash-chained, and a proxy mode is included.

What the research found

Mnestiq is a new open-source Python tool that records tamper-evident, Ed25519-signed hash-chained logs of AI agent runs, with a drop-in SDK for Anthropic and OpenAI clients and a proxy mode.

Claims and how well the evidence supports them

Records are hash-chained and sealed by Ed25519-signed Merkle checkpoints 99%
Install via pip install mnestiq with anthropic or openai extras 94%
A proxy mode records model calls without code changes 99%

Evidence

github.com https://github.com/mnestiq/mnestiq

Actionability rated medium.