Meet Gnito: Your AI. Your Machine. Your Rules.
What if you could run your own AI workspace locally, choose your AI provider, and keep control of your data?
That's what I'm building with Gnito.
Gnito is an open-source, local-first, self-hosted AI workspace built with React, TypeScript, Node.js and SQLite.
It doesn't bundle an AI model. Instead, it uses BYOK (Bring Your Own Key), allowing you to connect your preferred AI providers or local model endpoints.
What makes Gnito interesting?
- Multi-provider AI: Connect OpenAI, Gemini, Anthropic, Groq, Ollama, LM Studio and more.
- AI Chat: Streaming conversations, attachments, projects and chat history.
- Local Memory: Optional memory stored on your machine.
- Incognito Mode: Temporary conversations without persistent local history.
- Docker Terminal: Run AI-assisted commands inside an isolated environment.
- Web Search: Search the web with configurable search providers.
- Automation Center: Schedule AI tasks using cron or one-time schedules.
- GitHub Integration: Let AI interact with repositories through your GitHub token.
- Image Generation: Generate images through compatible providers.
- Encrypted Backups: Password-protected backups using AES-256-GCM.
Built with
- React + TypeScript
- Node.js + Express
- SQLite
- Vite
- Docker
Gnito runs locally and stores its workspace data under ~/.gnito/ by default.
External AI providers still process the prompts you send them, and their API usage may incur costs.
Try Gnito
Requirements: Node.js 20.19+.
git clone https://github.com/V-Dev-arch/gnito.git cd gnito npm install npm run dev Then open http://localhost:5173.
Docker is optional for most features but required for Terminal.
What's next?
Gnito is actively evolving. Some planned capabilities, including voice pipelines and MCP integration, are not implemented yet.
Contributions, suggestions and feedback are welcome!
GitHub: https://github.com/V-Dev-arch/gnito
Your AI. Your Machine. Your Rules.
