These docs describe Agent Behavior Lab as reusable software for safety and behavior evaluation of tool-using LLM agents.
- Repo Docs: documentation index
- Architecture: system layers and execution flow
- Session Guide: how to create and run evaluations
- Metrics and Analysis: result fields and built-in summaries
- Tool Variant System: baseline and alias tool representations
This repository does not include private result archives, private writing material, or preloaded fixtures. Add your own models, tools, personas, histories, judges, and sessions through the UI or REST API.