Turn learning materials into playable browser mini-games — self-hosted, OpenAI-compatible API, no account required.
Full documentation, demo video, screenshots, and bilingual (English / 中文) guide:
https://github.com/yh2072/edgameclaw#readme
Hosted product: ahafrog.com
Requires Python 3.10+ and Node.js 18+ (node on your PATH).
pip install edgameclawCreate .env in the directory where you run the server (or export env vars):
API_KEY=sk-your-openai-compatible-keyStart the app:
uvicorn edgameclaw.server:app --host 127.0.0.1 --port 8000Open http://127.0.0.1:8000 in your browser.
A small bundled Node service (for playing games) starts on port 3100 when that port is free. Set EDGAMECLAW_ENGINE_STATE_AUTO=0 if you run it yourself.
From source: git clone https://github.com/yh2072/edgameclaw → pip install -e . → see GitHub README for start.sh and .env.example.
| Variable | Notes |
|---|---|
API_KEY |
Required for generation (or paste in Studio UI only — browser-local). |
API_BASE_URL |
Default: OpenRouter. |
MODEL |
Default: google/gemini-3-flash-preview. |
EDGAMECLAW_HOME |
Optional: where to store courses and data (default: current working directory when installed via pip). |
- Source & full README: https://github.com/yh2072/edgameclaw
- License: AGPL-3.0 — see
LICENSE-AGPL-3.0in the repository - Citation / author: see GitHub README