About bolt.diy
bolt.diy is an open-source AI coding workspace, forked from StackBlitz's bolt.new, that generates and runs full-stack web applications in the browser. It began as a project by Cole Medin and is now maintained by the community under the MIT license. It suits developers and builders who want a prompt-to-app tool they can run themselves and point at whichever LLM they like.
Key benefits of bolt.diy include:
- Any model you want: Supports 22+ providers including OpenAI, Anthropic, Gemini, Groq, xAI, DeepSeek and Mistral, plus local models through Ollama and LM Studio.
- In-browser dev environment: Runs a full-stack Node.js environment powered by WebContainers.
- Interactive workbench: Includes a code editor, terminal, visual diff viewer and live preview.
- Flexible hosting: Run it with pnpm, Docker Compose, or as a desktop app for Windows, macOS and Linux.
- Free model option: Can use OpenRouter's Free Models Router, subject to OpenRouter's availability and rate limits.
If you are leaving Lovable, Base44 or v0, bolt.diy removes the subscription and the lock-in to one vendor's model. You pay your LLM provider directly, or nothing at all if you run local models. The trade-off is that you handle setup yourself: installing Node.js and pnpm or Docker, managing API keys, and securing remote deployments. The README notes that keys entered in the UI are stored in browser cookies, so it recommends HTTPS. Expect fewer hosted conveniences than a managed product, and output quality will depend on the model you choose.

