A TypeScript SDK for building agent-native apps with generative UI, shared state, and human-in-the-loop flows in React, Angular, Vue, and chat apps.
XiaoLuoLYG/GOD
A browser-based control room for running LLM agent societies, letting you pause, question, steer, and replay simulated towns step by step.
- Stars
- 1,228
- Forks
- 136
- Open issues
- 9
- Last commit
- 18 days ago
What is it?
GOD is a local control room for simulations of language-model agent societies. Instead of only watching a simulation run, you can pause at any step, scrub or auto-play a replay, ask one resident, a group, or the whole town a question, and inject instructions that agents read on their next turn.
The project runs a local FastAPI backend, a React/Vite browser UI, and an out-of-process agent runtime connected over a local WebSocket. A setup wizard configures any OpenAI-compatible model endpoint, a scenario, and agent profiles, then launches the experiment. Experiments, maps, and agent sets can be exported and imported as ZIP packs.
It ships two built-in experiments on a Tiled-based pixel town and a PKU campus map, plus a browser-only replay on GitHub Pages that needs no install or API key.
Key features
- 01Pause, scrub, and auto-play a live or recorded simulation by step.
- 02Send targeted questions to one agent, a group, or the whole town mid-run.
- 03Inject instructions that agents read on their next turn.
- 04Browser setup wizard for model, scenario, and agent generation.
- 05Map Studio and Agent Studio for editing maps and residents.
- 06Import and export experiment, map, and agent packs as ZIP files.
Use cases
- Inspecting how LLM agents behave under a scripted social scenario.
- Testing how interventions change a simulated community's outcomes step by step.
- Sharing a reproducible agent simulation as a downloadable pack.
- Prototyping custom maps and resident profiles for agent-based experiments.
Quick start
- 1
Clone the repository
git clone https://github.com/XiaoLuoLYG/GOD.git - 2
Enter the project directory
cd GOD - 3
Start GOD (macOS/Linux); the setup wizard opens in the browser
./scripts/god.sh start
Things to know
- Running locally requires an OpenAI-compatible model API key and endpoint, which the setup wizard asks for.
- The README also describes Windows startup via a PowerShell launcher, which is a separate command.
- The README states it is built for inspecting language-agent societies, not for claiming they are socially realistic.
Topics
FAQ
- What is GOD?
- A browser-based control room for running LLM agent societies, letting you pause, question, steer, and replay simulated towns step by step. GOD is a local control room for simulations of language-model agent societies. Instead of only watching a simulation run, you can pause at any step, scrub or auto-play a replay, ask one resident, a group, or the whole town a question, and inject instructions that agents read on their next turn.
- Is GOD free to use? What license does it have?
- GOD is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
- Is GOD still maintained?
- Actively maintained. The last commit to GOD was on Oct 2, 2026.
- What can I use GOD for?
- Inspecting how LLM agents behave under a scripted social scenario. Testing how interventions change a simulated community's outcomes step by step. Sharing a reproducible agent simulation as a downloadable pack. Prototyping custom maps and resident profiles for agent-based experiments.
- How do I install GOD?
- Clone the repository: git clone https://github.com/XiaoLuoLYG/GOD.git Enter the project directory: cd GOD Start GOD (macOS/Linux); the setup wizard opens in the browser: ./scripts/god.sh start
- Where can I get GOD?
- The source code and installation instructions are on GitHub: https://github.com/XiaoLuoLYG/GOD
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
A Python multi-agent framework for building, simulating and studying LLM-based agents, with tools, memory, and data generation support.
Open-source infrastructure for running AI-generated code in isolated cloud sandboxes, with JavaScript and Python SDKs.
A framework-agnostic Generative UI toolkit built around OpenUI Lang, a compact streaming language that lets LLMs produce renderable UI from your own component library.
A modular Agentic RAG tutorial and project built with LangGraph, featuring hierarchical indexing, conversation memory, and query clarification.
An open-source long-term memory layer for AI agents that stores and recalls information across sessions via a Python SDK, REST API, and MCP server.