AI to the People
What is WebLLM?
Section titled “What is WebLLM?”Install Browser Extension
One-time setup takes 30 seconds
Connect YOUR AI
Local models, Claude Pro, ChatGPT Plus, or company-provided
Every Website Can Use It
Your AI works everywhere, automatically
You Control Everything
Permissions, data, which AI to use per site
For Users: Your AI, Everywhere
Section titled “For Users: Your AI, Everywhere”🎯 Choose Your AI
- Download local models - Llama, Mistral, Phi-3 run in your browser
- Connect your subscriptions - Already paying for Claude Pro or ChatGPT Plus? Use them everywhere
- Use company models - Enterprise AI, private deployments, custom fine-tunes
- Switch anytime - Different AI for different sites or tasks
🔒 Control Your Data
- Local models = zero data sharing - Your data never leaves your device
- Per-site permissions - Approve or deny AI access for each website
- See what's being sent - Full transparency on every request
- Revoke access anytime - Change your mind? One click removal
💰 Stop Paying Twice
- Already have ChatGPT Plus? - Use it on every website, not just chat.openai.com
- Not paying anything? - Use free local models instead
- One subscription, infinite apps - Stop paying per-app for AI features
- Your choice, your wallet - Pick the AI that fits your budget
How It Actually Works
Section titled “How It Actually Works”You visit a website with AI features
A docs site with AI search, a writing app with smart suggestions, or a code editor
Website requests AI assistance
ai.generate({ prompt: "Summarize this article" })
WebLLM routes to YOUR chosen model
You already set your preference: local Llama for privacy, Claude for quality, or GPT-4 for reasoning
Results appear instantly
The website shows the response from YOUR AI, not theirs
Switch models and retry
Not happy with the result? Switch from Llama to Claude and regenerate in one click
For Developers: Enable the Future
Section titled “For Developers: Enable the Future”Free Infrastructure
- Zero API costs
- No key management
- No billing system
- No secrets storage
Universal Compatibility
- Works with any model
- Automatic updates
- Future-proof code
- No vendor lock-in
Privacy Built-in
- GDPR/HIPAA ready
- No data storage
- User-controlled
- Audit-friendly
Smart Routing
- Auto-select best model
- Graceful fallbacks
- Quality guarantees
- Task optimization
Users Want This
- Growing adoption
- Competitive advantage
- User trust
- Lower churn
3 Lines of Code
const ai = window.ai;
const result = await ai.generate({
prompt: "Your request"
});The Problem We’re Solving
Section titled “The Problem We’re Solving”❌ The Current Reality
- Every app has its own AI - You pay separately for Notion AI, Grammarly AI, GitHub Copilot, etc.
- No control - Can't choose your model, can't see what data is sent, can't revoke access
- Developers build everything - API infrastructure, billing, rate limiting, model management
- Lock-in everywhere - Each app picks one model, you're stuck with it
✅ The WebLLM Solution
- Your AI, your control - One subscription (or local models), works everywhere
- Full transparency - See every request, approve or deny, switch models anytime
- Developers ship features - 3 lines of code, no infrastructure, no costs
- Universal compatibility - Works with any model, new models added automatically
WebLLM is like ad blockers, but for AI.Users install once and take control. Websites work better. Everyone wins.
Get Started
Section titled “Get Started”🧑 For Users
- Download Extension - Install from Chrome Web Store (30 seconds)
- Connect Your AI - Add API keys or download local models
- Browse Normally - Visit any site with AI features, approve access, enjoy
👨💻 For Developers
- Read Docs - 5-minute quickstart guide
- Add 3 Lines - window.ai.generate() is all you need
- Ship Features - Deploy today, zero infrastructure
🏢 For Enterprises
- Self-host or cloud - Deploy WebLLM server in your infrastructure
- Connect private models - Azure OpenAI, AWS Bedrock, custom endpoints
- Roll out to org - Centralized control, user privacy, compliance ready
Supported Providers
Section titled “Supported Providers”Supported Providers
And many more through OpenAI-compatible APIs and Vercel AI SDK
16+ providers supported out of the box - Anthropic, OpenAI, Google, Azure, DeepSeek, Mistral, Groq, Cohere, Together.ai, OpenRouter, Ollama, LM Studio, and more. New providers added regularly.
Why This Matters
Section titled “Why This Matters”“The web won because it was open. Email won because it was federated. AI should work the same way - no walled gardens, no lock-in, just protocols.”
User Sovereignty
You choose your AI provider, not each website
Developer Freedom
Build AI features without infrastructure costs
Privacy by Design
Local processing is always an option
Open Standards
Moving toward W3C browser API standardization
Frequently Asked Questions
Section titled “Frequently Asked Questions”Do I need to install an extension?
Currently yes - WebLLM is a Chrome extension polyfill while we validate the API design. We're working toward native browser integration, but the extension is fully functional today.
What if a website doesn't support WebLLM?
The extension only activates when websites request AI capabilities through the WebLLM API. Traditional sites work exactly as before. As more developers adopt WebLLM, more sites will support it.
Is my data safe?
You control everything. Use local models for zero data sharing, review every request before it's sent, grant per-site permissions, and revoke access anytime. WebLLM itself stores no data externally.
How do local models work?
WebLLM downloads model files to IndexedDB and runs inference using WebGPU (GPU) or WebAssembly (CPU). Models run entirely in your browser - no network requests after download.
Can I use my existing ChatGPT/Claude subscription?
Yes! Add your API key once and use it across all WebLLM-enabled websites. Stop paying per-app for AI features you've already subscribed to.
Is this production-ready?
Yes. The extension and SDK are stable and actively used. The API design is being validated before proposing to W3C for standardization.
Ready to Join the Movement?
Whether you’re a user taking control or a developer shipping features, WebLLM makes AI better for everyone.