An AI Native Guandan Card Game
A semi-entertainment, semi-research AI Native Guandan card game project, supporting mixed matches between AI agents and human players.
clawguandan is a semi-entertainment, semi-research AI Native Guandan card game project. It implements Guandan, one of the most popular card game formats in China.
With this project, you can easily run mixed matches between AI players and human players. It can be used both for casual fun and for research: observing and comparing strategy, collaboration, and technical progress across different LLMs in real gameplay environments.
- Full implementation of core Guandan game flow and rule logic
- HTTP API-based client/server architecture for both local and remote deployment
- A polished Web UI and CLI, enabling seamless human-AI interaction
- Support for natural language AI integration via OpenClaw / Hermes Skills
Available for:
- Linux (aarch64 & x86_64)
- macOS (Apple Silicon & Intel)
- Windows WSL2
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/mikewei/clawguandan/releases/latest/download/clawguandan-installer.sh | sh
npm install -g @mikewei-labs/clawguandan@latest
cargo install clawguandan
hermes skills install --force mikewei/clawguandan/hermes-skills/guandan
npx clawhub@latest install guandan
- Start the server:
clawguandan server start- Add AI players :
If you prefer hermes:
clawguandan bot llm-bot --players 3 --default-script hermesOr you can also use openclaw:
clawguandan bot llm-bot --players 3 --default-script openclaw- Let human players join:
Open http://127.0.0.1:22222 in your browser to enter the game UI. (Replace the IP with the host where the CLI is running)
Once skill installed, you can just prompt "Let's play guandan together! Create a table with three AI bots." to start the game.