XiaoLuoLYG/GOD

What it solves

GOD (Govern, Observe, Direct) is a tool for inspecting and controlling language-agent societies. Unlike traditional agent simulations that only allow users to watch, GOD provides a "control room" to actively manage, intervene in, and analyze the behavior of multiple AI agents interacting in a simulated environment.

How it works

GOD operates as a local-first system consisting of a React/Vite frontend (the Control Room), a FastAPI backend, and an out-of-process agent runtime. It separates the simulation into three components: Experiments (the setup including maps and agents), Replays (the viewable results), and Runtime state (local data like SQLite stores). Users can configure the system via a browser-based setup wizard and connect it to any OpenAI-compatible LLM endpoint.

Who it’s for

It is designed for researchers and developers who want to study agent societies, test hypotheses about agent interactions, and create reproducible agent-based experiments without needing to write extensive code.

Highlights

• Real-time Intervention: Inject instructions into the next step of a simulation to see how agents respond.
• Replay Control: Scrub, pause, fast-forward, or jump through recorded or live simulation steps.
• Targeted Communication: Send natural-language questions to specific agents, groups, or the entire town mid-run.
• No-Code Setup: A browser wizard for configuring models, scenarios, and generating agent profiles and step plans.
• Studio Tools: Includes an Agent Studio for editing resident profiles and a Map Studio for generating and calibrating custom maps.
• Portable Packs: Import and export experiments, maps, and agent packs via ZIP files.