bolt.diy

bolt.diy is an open-source fork of Bolt.new, positioned as "the open-source AI developer workspace that works with any LLM you want." It allows users to freely choose their preferred LLM for each prompt, enabling production-grade builds of full-stack web applications in the browser. The project was initially started by Cole Medin and later grew into a large-scale community collaboration project, with the goal of building the best open-source AI programming assistant.

Core Purpose

Users can generate, run, edit, and deploy full-stack web applications through prompts. Unlike tools tied to a single model, bolt.diy's core selling point is freedom of model choice: it can connect to any provider supported by the Vercel AI SDK, and also supports user extension.

Main Capabilities

• In-browser web development environment: Based on WebContainers, it provides a complete full-stack Node.js development environment.
• 22+ AI provider support: Cloud providers cover Anthropic, OpenAI, Google, DeepSeek, Groq, xAI, Mistral, and more; local providers support Ollama and LM Studio. Among them, OpenRouter, Anthropic, OpenAI, and Google are enabled by default.
• Interactive workbench: Integrates a code editor, terminal, visual diff viewer, and live preview. It opens by default when chat starts, and can be reopened after closing.
• Starter Setup: Explicitly requested frameworks will match the corresponding starter in the code; starter files include detected install commands and dev server commands, and are prepared before the first custom request.
• Prompt library: System prompts can be selected under Settings → Workspace → Features, including Simple Sal for concise guidance and Tiny Tim for smaller models.
• Project workflow: Supports Git clone and push, snapshot restore, file conflict locking, and ZIP export.
• Deployment: One-click deployment to Netlify, Vercel, and GitHub Pages.
• Database and visualization: Integrates Supabase, data visualization charts, and Model Context Protocol (MCP) tools.
• Expanded AI and tool ecosystem: 22+ cloud and local LLM providers, MCP integration, and voice prompts.
• Enhanced code and project workflow: Visual diff inspection, file conflict locking, codebase search, Git integration, and snapshot restore.
• Cross-platform and deployment: Native Electron desktop app, Expo (React Native) support, and direct deployment to Netlify, Vercel, GitHub Pages.
• Data and backend integration: Built-in Supabase management, interactive charts, and batch chat operations.
• Prompt and connector control: MCP tools can be selected from the prompt attachment menu; manage service connections such as GitHub, GitLab, and Netlify under Settings → Integrations → Connectors.

Usage

Runtime Requirements

• Node.js >= 22.12.0
• pnpm >= 9.15.9 (can be installed via npm install -g pnpm installation)

Local Run

1. Clone and install
2. Run the development server
3. Open in the browser http://localhost:5173

Docker Run

• Configure environment variables:cp .env.example .env.local
• Build and run:
  • Development mode (hot reload):docker compose --profile development up
  • Production build:pnpm run dockerbuild:prod, then docker compose --profile production up

Desktop App (Electron)

• Prebuilt desktop binaries: Download Windows, macOS, and Linux versions from Latest Releases.
• Build from source:pnpm electron:build:dist

Configure API Keys and Providers

Option A: Configure in the UI (recommended) Click the settings icon (⚙️) in the sidebar → Providers. Enter an API key or configure a local endpoint (Ollama, LM Studio, OpenAI-compatible endpoint). Keys are validated in real time and securely stored in browser cookies.

Option B: Environment variables Set .env.example Copy as .env.local, and add a key or local endpoint.

Use Cases

Suitable for developers who want to quickly build a full-stack web application in the browser and need the flexibility to choose different LLMs (especially those who want to use local models or specific cloud models). Also suitable for a complete project workflow that requires Git version management, one-click deployment, a Supabase backend, and an MCP toolchain.

Limitations and Notes

• The project is currently in maintenance status: the "Looking Ahead" section states that the ongoing work is mainly bug fixes, dependency updates, and reliability fixes, and no new features are planned.
• It is open source under the MIT License, but it uses the WebContainers API, which has specific commercial licensing terms for enterprise deployment.
• Documentation, community forums, FAQ, and troubleshooting are officially provided by the project; common errors and installation fixes are recorded in the FAQ and official documentation.
• Contributions and community: contributions are welcome, and the project provides a Contributing Guide and Project Management Guide.
• Pricing, specific charging plans, platform limitations, etc. are not yet provided in the given materials.