Agent observability and replay tooling for AI safety & interpretability research. - View it on GitHub