Community
A free, open-source AI chat app you can run locally in minutes with your own API key. It is the real architecture, deliberately kept small.
Quickstart
Requires Node 22+. No database, no Docker.
git clone https://github.com/startoneai/startone-community.git
cd startone-community
npm install
cp .env.example .env.local
npm run dev
Open http://localhost:3000.
Configuration
Set keys in .env.local. All are optional — add only the ones you want.
| Variable | Enables |
|---|---|
OPENAI_API_KEY | GPT models |
ANTHROPIC_API_KEY | Claude models |
GOOGLE_GENERATIVE_AI_API_KEY | Gemini models |
ZAI_API_KEY | GLM models |
DEEPSEEK_API_KEY | DeepSeek models |
DASHSCOPE_API_KEY | Qwen models |
FAL_KEY | The image-generation demo (fal.ai) |
Keys are read from the environment only. There is no database, and nothing is sent anywhere except directly to the provider you configured.
Models available
Six chat models, selectable in the picker: gpt-5.5, claude-opus-4-8, gemini-3.1-pro, glm-5.2, deepseek-chat, and qwen-max. Each needs its provider's key set; without one, that model returns a dry-run reply.
Scripts
| Command | What it does |
|---|---|
npm run dev | Start the dev server |
npm run build | Production build |
npm run typecheck | TypeScript, no emit |
npm test | Run the test suite |
npm run gate | Lint + typecheck + test — run before committing |
Limits — what Community deliberately does not do
These are design decisions, not missing features. Community exists to show you the architecture honestly, not to be a free version of a paid product.
- The image demo is watermarked and capped at 512×512. It proves the pipeline works; it is not for production.
- No video or audio generation.
- No gallery. Generations are not stored or browsable.
- No authentication. Single-user, local.
- No database. Chat history is session-only — refresh and it is gone.
- No billing.
When you need any of the above, see Developer — Web.
License
MIT. Free to use, modify, and build on. The StartOne name and logo are a separate matter — see Licensing.