WebLLM
AI Just Got Standardized
A Boring Standard That Changes Everything
WebLLM is an open protocol that makes AI universal. Users choose their provider. Developers skip the infrastructure. The web gets AI everywhere.
AI should be a utility.
Instead it's a mess.
Users pay 5x
Same model, different apps, multiple bills
Developers rebuild
Identical infrastructure, over and over
Wikipedia can't afford
What individual users pay for
3 companies control
How we get access to AI
10 Reasons whyBrowser-Native AIis inevitable
End the AI Tax
Stop paying multiple times for the same technology
Users pay for GPT-4 five times across different apps. One subscription should work everywhere — like having one internet bill, not paying each website separately.
After: Pay once → Use everywhere
Total: $75/mo
Same GPT-4 model
Democratize AI Access
Everyone deserves AI, not just the venture-funded
AI shouldn't be a luxury for funded startups. When users bring their own AI, every website can be intelligent.
All AI-enabled now
We already own the hardware
Make use of the computers people already have
Millions of powerful GPUs and M-series Macs sit idle while their owners pay for cloud AI. That's backwards.
• Privacy bonus: Your data never leaves your device
• Speed bonus: Often faster than API calls
Your hardware finally wakes up and works for you
RTX 4090
82 TFLOPS AI
M3 Max
40 TFLOPS AI
Eliminate Infrastructure Redundancy
Stop rebuilding the same plumbing
Every company spends weeks building identical infrastructure: auth, billing, rate limiting, error handling. Collective waste in the millions of developer hours.
Ship features, not infrastructure
Before
4 weeks
After
navigator.llm4 lines
Future-Proof Development
Write once, evolve forever
Stop hardcoding model names that become obsolete in 6 months. Describe what you need, let the system evolve.
When GPT-5 or Claude 4 launches, your code automatically uses it. No refactoring. No deployment.
Standards outlive products
Hardcoded approach:
Constant refactoring
Standard approach:
navigator.llmAuto-upgrades to best model
Enable True Client-Side AI
AI at the speed of thought
Every AI call shouldn't require a round-trip to your server, which often just relay requests to other AI services. Tools should execute where the data lives — in the browser.
• Instant tools: Manipulate DOM, update state immediately
• Real-time: Forms that complete as you think
The backend was always unnecessary overhead
Traditional:
~700ms round trip
With WebLLM:
~10ms instant
Guarantee Privacy by Design
Data sovereignty for users
Privacy shouldn't depend on corporate promises. When AI runs locally, sensitive data never leaves your control.
Privacy by architecture, not by policy
Medical
Local only
Financial
Local only
Work docs
Local only
Personal
Local only
Restore User Agency
Users control their tools, not vice versa
You shouldn't need permission to use your preferred AI. Your tools should adapt to you, not lock you in.
• Switch anytime: No app breakage when you change
• Mix and match: Different AI for different tasks
Your preferences, your choice, everywhere
Your AI preferences:
Claude
Primary choice
GPT-4
Backup
Gemini
Available
Works on every website
Level the Playing Field
Small teams can compete with giants
Infrastructure advantage disappears when users bring their own AI. Now it's about ideas, not capital.
• Open source: Compete with commercial products
• Students: Build without credit cards
David and Goliath use the same AI
BigCorp
1000 devs
Solo Dev
1 person
Same AI capabilities
Standards Always Win
Proprietary products become universal utilities
The web thrives on open standards. AI should be no different — universal, interoperable, and user-controlled.
Standards boring. Standardization revolutionary.
<video>fetch()WebGLnavigator.llm527
Developers building
10,241
Users connected
142
Commits this week
This isn't a product launch.
It's a protocol shift.
What Developers Are Saying
See what the community thinks about our LLM Gateway
Works with your AI Providers
Plus Chrome's built-in Gemini Nano, and many more through OpenAI-compatible APIs
Significant sponsors, current and past (smaller logos)
They already use WebLLM
Discover how developers are building intelligent applications with browser-native AI.
Inspire Cards
AI-powered card interpretation for brainstorming, idea generation, card games, and activities like tarot reading. Let users explore and discuss cards with conversational AI.
The Roadmap
standards always win
Common Questions & Answers
Find out all the essential details about WebLLM and how it can serve your needs.
What is WebLLM and how does it work?
WebLLM is an open protocol that brings AI directly to your browser. Instead of each app having its own AI integration, WebLLM provides a standard SDK that lets users bring their own AI provider to any website.
Do I need to pay for multiple AI subscriptions?
No! That's exactly the problem WebLLM solves. Instead of paying $20/month to ChatGPT, $10 to Notion AI, $30 to Grammarly — all using the same underlying model — you pay once and use it everywhere through WebLLM.
Is my data private when using WebLLM?
Privacy is built into the architecture. You control which AI provider handles your data. For maximum privacy, you can run models locally on your own hardware — your data never leaves your device.
Does WebLLM work with Chrome's built-in AI?
Yes! Chrome 138+ includes Gemini Nano — free, on-device AI for desktop browsers. WebLLM automatically uses it when available, with graceful fallback to other providers. Users with Chrome AI get free local AI; everyone else uses their configured provider.
How do I get started as a developer?
It's just a few lines of code. Install the WebLLM SDK, create a session, and start making requests. No API keys to manage, no billing to set up — WebLLM handles it all through the user's configured provider.
Which AI providers are supported?
WebLLM supports 30+ providers including OpenAI, Anthropic, Google, Mistral, Groq, Together.ai, OpenRouter, Chrome's built-in Gemini Nano, and local options like Ollama and LM Studio. Users choose their preferred provider and can switch anytime.
Is WebLLM free to use?
WebLLM itself is completely free and open source. Users pay only for their chosen AI provider (or nothing if using Chrome AI or local models). Developers can add AI features to their apps without any infrastructure costs.
Help Build the Web
AI as a utility needs builders, users, advocates, and believers.
How will you help build the grid?
It's easier than you think.
→ See Quick Start (4 lines of code)
→ View Examples
→ Join Discord
Make AI work like it should.
→ Add to Chrome
→ Configure Your AI
→ Browse Normally
Share what's possible.
→ Tweet Template
→ Email Template
→ Success Stories
Accelerate the future.
This is a community effort, privately funded.
We believe it has the potential to disrupt the market and shift value to small players and individual users.
Utilities aren't built by companies.
They're built by communities.
Are you in?