WebInfer

Web Inference Protocol

One protocol. Any provider. User-controlled.

Let's make AI work as it should.

How It Works

Apps request AI. The protocol delivers it.

Client

@webinfer/client

A lightweight SDK that auto-connects to available AI sources. CallgenerateText(),generateImage(), orgenerateSpeech().

What it does:

  • +Simple API for text, image, audio, speech
  • +Auto-connects to user's AI sources
  • +Handles streaming responses
  • +Tool calls for data retrieval & UI

You don't manage:

  • API keys (users bring their own)
  • Provider selection (automatic)
  • Server infrastructure (none needed)
Server

@webinfer/server

The brain that handles requests. Routes to the best provider—local or cloud—based on task, preferences, and availability.

What it does:

  • +Manages provider connections & API keys
  • +Routes requests to optimal provider
  • +Handles fallbacks when providers fail
  • +Stores user preferences & history

Runs in:

  • Browser extension
  • Native desktop app
  • Cloud gateway
App calls client
Client talks to server
Server routes to provider

For Developers: The Client is All You Need

Import the client. Call a function. Ship AI features—no API keys, no server costs.

Build AI Features in Minutes

Just import @webinfer/client. The client connects to Chrome's built-in Gemini Nano, the user's OpenAI key, Ollama, or any configured provider—automatically.

+

Zero API Keys

Users bring their own.

+

Zero Server Costs

AI runs on user's side.

+

25+ Providers

One API handles all.

// That's it. Really.
import { generateText } from '@webinfer/client'
const result = await generateText({ prompt: "Hello!" })

Want to move even faster? Check out AI Blocks pre-built React components for chat, speech, images, and more. Copy, paste, ship.

Where does inference happen?

Users connect their AI sources. The protocol routes each request to appropriate models and providers.

Someone browsing the web on Chrome desktop

Gemini Nano

Chrome's built-in AI. Free, on-device, private.

Local Models

Ollama, LM Studio, or other local runners.

Commercial APIs

OpenAI, Anthropic, Google—your own API keys.

WebInfer Cloud

One subscription, 200+ models.

Corporate Gateway

Company-provided AI. Paid by employer.

Courtesy Credits

Free credits from app developers. Rate-limited.

Private Server

Self-hosted gateway. Access from anywhere.

Casual User has access to 2 inference sources. The server automatically picks the best one for each request.

Run Anywhere

The server runs wherever you need it. Same protocol, different packaging.

Browser Extension

Chrome & Chromium

Install once, works everywhere. Zero configuration. Perfect for getting started.

  • +Zero setup
  • +Auto-updates
  • +Works with any website
Install Extension

Native App

macOS, Windows, Linux

Works in every browser—Firefox, Safari, Edge, Arc. Run Python and C++ models locally.

  • +All browsers supported
  • +System notifications
  • +Python & C++ models
Download

CLI Daemon

Node.js

Lightweight daemon for developers. Run on your dev machine, Raspberry Pi, or home server.

  • +npx webinfer daemon
  • +Scriptable
  • +Docker-ready
npm install

WebInfer Cloud

Managed service

One subscription, 200+ models. No servers to manage. Works on any device.

  • +200+ models
  • +No API keys needed
  • +Cross-device
Learn More

Self-hosted Gateway

Your infrastructure

Deploy on your servers. Issue tokens, enforce quotas, monitor usage.

  • +Access tokens
  • +Usage quotas
  • +Federation
Deploy Guide
Optional Account Sync

Sign in to sync settings across devices. Or don't—everything works offline. No account required.

25+ Providers, One API

Connect your existing providers or discover new ones. Mix local with cloud. Switch without changing code.

Cloud APIs

Bring your own API keys

OpenAI
Anthropic
Google AI
Mistral
DeepSeek
Groq
Cohere
Azure

Local Runners

Run models on your hardware

Ollama
LM Studio
llama.cpp
Transformers.js

Browser-Native

Free, on-device, zero setup

Chrome Gemini Nano

Gateways

Access multiple providers

OpenRouter
Portkey
Together AI
Fireworks

Free AI, Built Into Chrome

Chrome's Gemini Nano runs entirely on your device. No API key, no cost, complete privacy. WebInfer uses it automatically when available.

Node Federation

Your AI, your infrastructure, your rules. Nodes connect and fall back to each other automatically.

Extension
Home Server
Cloud Gateway
AI Provider

Resilience

Home server offline? Automatic fallback to cloud.

Cost Control

Local models first, cloud only when needed.

Privacy

Sensitive requests stay on your hardware.

No Single Point

Distributed by design. Add nodes as needed.

Server Discovery

LikeMCP servers, WebInfer servers can be discovered and added automatically. A standardized endpoint advertises capabilities.

Well-Known Endpoint

/.well-known/webinfer.json

Every WebInfer server exposes a discovery endpoint. Clients can fetch server capabilities, available models, and authentication methods automatically.

+ Machine-readable format
+ No manual configuration
+ Dynamic capability detection
Response Format
{
  "$schema": "webinfer.com/schemas/...",
  "version": "1.0",
  "protocol": "webinfer",
  "provider": {
    "id": "my-gateway",
    "name": "My Gateway",
    "type": "gateway"
  },
  "server": { "baseUrl": "..." },
  "config": {
    "fields": [
      { "key": "apiKey", "type": "password" }
    ]
  },
  "auth": { "methods": ["bearer"] },
  "capabilities": ["chat", "image"],
  "models": [{ "id": "gpt-4o" }],
  "federation": { "supported": true }
}

Adding a Server

1

Enter URL

Paste the server URL. Discovery happens automatically.

2

Authenticate

Add your access token or API key if required.

3

Ready

Server models appear automatically. Set priority as needed.

Inspired by Model Context Protocol (MCP) server discovery patterns.

Don't Want to Host a Node?

WebInfer Cloud provides managed infrastructure so you don't have to run anything yourself.

WebInfer Cloud

Access 200+ models from a single subscription. Use it across all apps that support the protocol.

Learn More

Courtesy Compute

Give your users free AI credits. No setup required on their end. Enforce quotas to prevent abuse.

For Developers

"Isn't this just another provider trying to sell me a plan?"

Fair question. Yes, we offer a cloud service. But here's the thing: we want competitors.

The protocol is designed to make inference portable. We hope many players will emerge—specialized models, domain-specific services, regional providers, budget options. That's the point.

Our cloud keeps a small platform fee to fund development of the protocol, the core libraries, and the fun parts. But you can always self-host, use your own API keys, or connect to competing services.

We're building plumbing, not a walled garden.

What It Enables

A standard protocol unlocks capabilities that fragmented integrations can't.

Pay once, use everywhere

Stop paying $20 to ChatGPT, $10 to Notion AI, $30 to Grammarly—all using the same models.

True privacy

Your data never touches app servers. It flows directly from your browser to your chosen AI provider.

Free infrastructure

Add AI to your app without managing API keys, billing, or infrastructure. Users bring their own compute.

Provider freedom

Switch from OpenAI to Anthropic to local models without changing a line of code. Your choice, not the app's.

Local-first

Run models on your own hardware. Chrome's built-in Gemini Nano, Ollama, LM Studio—all work seamlessly.

Future-proof

Apps built on the protocol automatically gain access to new models and providers as they launch.

Custom AI implementation

  • Developer manages API keys
  • Provider locked by developer
  • Data flows through app servers
  • Developer pays (or passes cost to user)
  • 1-2 providers, hardcoded
  • Local models need separate implementation

With the Protocol

  • User configures once, works everywhere
  • User chooses their provider
  • Data goes direct to provider
  • User pays their own provider
  • 25+ providers automatically
  • Local models use the same API

Questions

What stage is this project in?

Experimental. The protocol and SDK are functional and used by early adopters, but APIs may change. We're iterating based on real-world feedback.

How is this different from calling OpenAI directly?

Users bring their own provider—you don't manage API keys. Privacy is built-in since you never see user data. And your app works with 25+ providers automatically.

What about Chrome's built-in AI?

WebInfer integrates with Chrome's Gemini Nano when available. It provides a unified API regardless of whether AI runs locally or in the cloud.

Is my data private?

Yes. Data flows directly from the browser to the user's chosen provider. WebInfer never sees prompt content or responses.

Can I use this in production?

Yes, with caution. The protocol is stabilizing but may have breaking changes. Many sites are already using it successfully.

Try It Out

Install the extension to experience the protocol, or read the docs to integrate it into your app.