Your AI workspace lives in your browser.

ai.diy is a local-first, bring-your-own-key chat workspace. Chats, Canvas, knowledge, and settings stay in your browser by default while you choose from 20+ cloud and local provider integrations. Add web search, MCP tools, Python, and approved agents without giving up the workspace. No persistent provider keys are required on the server.

Open workspaceStar on GitHub
  • LOCAL-FIRST
  • 20+ PROVIDERS
  • MIT LICENSED
  • NO PERSISTENT LLM KEYS
ai.diy — workspace

Workspace demo · click to expand

Your keys stay under your control.

Provider credentials stay in your browser's storage and are relayed per request. Threads, Canvas, memory, and knowledge persist in IndexedDB. The Node server is a relay, not a persistent provider-key store.

Stored in the browser

Provider credentials stay in browser storage and are relayed per request. They are not required as persistent server secrets.

Browser is the database

Threads, Canvas, memory, knowledge, usage events, and website projects persist in IndexedDB.

Any model mid-thread

20+ cloud and local providers. Switch without moving the workspace or losing context.

Bring any model into the same thread.

Connect 20+ cloud and local providers, then switch models without moving your workspace or copying context between apps.

  • OpenAIOpenAI
  • GrokGrok
  • GeminiGemini
  • ClaudeClaude
  • OpenRouterOpenRouter
  • DeepSeekDeepSeek
  • OllamaOllama

ALSO AVAILABLE: ANTHROPIC · GROQ · MISTRAL · BEDROCK · AZURE · VERTEX · TOGETHER · HUGGING FACE · LM STUDIO · CUSTOM ENDPOINTS

Private by default. Capable by design.

  • Local code review

    Refactor and review with BYOK or Ollama. Diffs stay in the browser; nothing is uploaded to a vendor workspace.

  • Private document RAG

    Search your notes and PDFs with on-device embeddings. The knowledge base never leaves this machine.

  • Interactive Canvas

    Preview HTML, Python files, and code artifacts beside the thread — live outputs without cloud lock-in.

Tools, storage, and deployment on your terms.

Active laneKeyless searchWebsite presetsAgent Mode

Search, skills, MCP, subagents.

DuckDuckGo plus Firecrawl and Parallel MCP ship keyless. URL fetch, calculator, browser Python, on-device knowledge RAG, memory, remote MCP, slash skills, approved subagents, and experimental website presets. Enabled tools have their own service boundaries.

Search

DuckDuckGo · Firecrawl · Parallel

Runtime

Browser Pyodide · Web Speech

Agents

Skills · Subagents · MCP

Self-host in one command.

Node production build, Docker Compose, or a Vercel preview. Click the terminal to copy.

Deploy to Vercel

Frequently asked questions.

Understand the ownership model, supported AI providers, browser storage, and self-hosting path before you open the workspace.

What is ai.diy?

ai.diy is an open-source, self-hosted AI workspace for chat, research, tools, Canvas artifacts, memory, and local knowledge. It uses bring-your-own-key (BYOK) access so you choose the provider and model instead of being locked into one hosted AI service.

Are my AI provider keys stored on the ai.diy server?

No server-side LLM credentials are required. Provider keys are kept in your browser and relayed per request to the provider you select. A hosted instance operator can observe traffic in transit, so self-host ai.diy when you need control over the infrastructure and network boundary.

Where does ai.diy store chats and documents?

Chats, files, Canvas artifacts, memory, on-device knowledge-base chunks, usage events, and preview sessions persist in your browser through IndexedDB and localStorage. Optional S3, WebDAV, or Google Drive backups are client-side features that you enable yourself.

Which AI providers work with ai.diy?

ai.diy supports 20+ integrations including OpenAI, ChatGPT subscription, Anthropic, Google Gemini, Groq, Cerebras, Fireworks, Perplexity, Cohere, OpenRouter, DeepSeek, xAI, Ollama, Mistral, Hugging Face, Amazon Bedrock, Azure, Vertex, Vercel Gateway, Together, LM Studio, and custom OpenAI-compatible endpoints.

Can I self-host ai.diy?

Yes. Run the production build on a standard Node.js server or use Docker Compose. The server acts as a request relay and does not need provider API keys in environment variables. You can open the hosted demo first or deploy the MIT-licensed source code on infrastructure you control.

Does ai.diy replace my AI provider or pay for model usage?

No. ai.diy is the workspace layer. You bring authorized provider keys or connect local models such as Ollama and LM Studio. Provider pricing, quotas, availability, and data policies remain controlled by each provider, and any usage charges are yours.

Your models. Your data. Your machine.

Open the workspace, drop in a key, and start thinking. Or clone the repo and self-host in one command.

Open workspaceView on GitHub

20+ PROVIDERSNO PERSISTENT LLM KEYSINDEXEDDB PERSISTENCEMIT LICENSED