AATF – An open spec for recording why AI agents make decisions

Just trying to figure if this is something different from what traditional logging offers. Plus on tooling option how more effective is it wrt to tools like mlflow.