Skip to content

README.md

Latest commit

 

History

History
299 lines (226 loc) · 11.4 KB

README.md

File metadata and controls

299 lines (226 loc) · 11.4 KB

🌉 web-model-bridge

Bridge web AI models through OpenAI-compatible API

Use Claude, ChatGPT, DeepSeek, and 8 more AI models — completely free, zero API tokens.

License: MIT Node.js TypeScript Tests

Quick Start · Supported Models · Configuration · API Reference · Contributing

What is this?

web-model-bridge is a standalone HTTP service that lets any AI tool use web-based AI models through their free browser interfaces. It acts as a bridge between AI tools (OpenClaw, Claude Code, Cursor, etc.) and web AI platforms (Claude, ChatGPT, DeepSeek, etc.).

Your AI Tool  →  web-model-bridge  →  Browser  →  Web AI Model
(OpenClaw)       (localhost:3456)     (Chrome)    (claude.ai)

How it works: You log into AI websites once through a Dashboard. The bridge then uses your browser session to forward API requests — no API keys, no tokens, no cost.

Why web-model-bridge?

Compared to alternatives like gpt4free (66K stars), CLIProxyAPI (23K stars), and chat2api (3.4K stars):

web-model-bridge gpt4free CLIProxyAPI chat2api
Approach Real browser automation Reverse-engineered APIs CLI OAuth proxy Token simulation
Anti-blocking Strongest — real browser fingerprint Weak — APIs break often Medium Weak — Cloudflare blocks
Cost Free — web free tier only Free Needs $20-100/mo subscription Free
Platforms 11 platforms, 16 models Varies (unstable) 4-5 platforms ChatGPT only
API formats OpenAI + Anthropic OpenAI only OpenAI + Anthropic OpenAI only
Language TypeScript (Node.js native) Python Go Python

Core advantages:

  1. Strongest anti-blocking — Uses Playwright real browser, websites can't distinguish from normal browsing
  2. Truly free — Only needs free web accounts, no paid subscriptions required
  3. Widest coverage — 11 platforms (international + Chinese), 16 models in one bridge
  4. Dual API format — Both /v1/chat/completions (OpenAI) and /v1/messages (Anthropic), works with every AI tool
  5. Node.js ecosystem — TypeScript native, npx web-model-bridge zero-install startup

Features

Feature Description
🔌 11 Providers Claude, ChatGPT, DeepSeek, Kimi, Qwen, GLM, Grok, Gemini, Perplexity, Doubao, Xiaomimo
🔄 Dual API Format OpenAI (/v1/chat/completions) + Anthropic (/v1/messages)
🖥️ Web Dashboard Visual management — login, status, one-click API URL copy
🚀 One Command npx web-model-bridge — auto environment check, auto open Dashboard
🔒 Secure Localhost-only by default, optional Bearer token, browser-isolated cookies
💻 Cross-Platform macOS, Linux, Windows
🎯 Zero Config Works out of the box, optional YAML config for customization

Quick Start

1. Start the bridge

npx web-model-bridge

This will:

  • ✓ Check your environment (Node.js, Chrome)
  • ✓ Start HTTP server on port 3456
  • ✓ Open Dashboard in your browser

2. Login to AI providers

In the Dashboard (http://localhost:3456), click Login next to any provider. A browser window opens — log in as you normally would. Done.

3. Connect your AI tool

OpenClaw — Add to ~/.openclaw/openclaw.json:

{
  "models": {
    "mode": "merge",
    "providers": {
      "webmodel": {
        "baseUrl": "http://127.0.0.1:3456/v1",
        "apiKey": "not-needed",
        "api": "openai-completions",
        "models": [
          { "id": "deepseek-web/deepseek-v4", "name": "DeepSeek V4 (Free)", "contextWindow": 128000, "maxTokens": 8192 }
        ]
      }
    }
  }
}

Claude Code:

export ANTHROPIC_BASE_URL="http://localhost:3456"
export ANTHROPIC_API_KEY="not-needed"
claude

Cursor: Settings → Models → Override OpenAI Base URL → http://localhost:3456/v1

Any OpenAI-compatible tool: Set base URL to http://localhost:3456/v1

Supported Models

Model ID Name Context Platform
claude-web/claude-sonnet-4-6 Claude Sonnet 4.6 1M claude.ai
claude-web/claude-haiku-4-5 Claude Haiku 4.5 200K claude.ai
chatgpt-web/gpt-5.3 GPT-5.3 128K chatgpt.com
chatgpt-web/gpt-5.4-mini GPT-5.4 Mini 128K chatgpt.com
deepseek-web/deepseek-v4 DeepSeek V4 128K chat.deepseek.com
deepseek-web/deepseek-v4-reasoner DeepSeek V4 Reasoner 128K chat.deepseek.com
kimi-web/kimi-k2.5 Kimi K2.5 256K kimi.moonshot.cn
qwen-web/qwen-3.5-plus Qwen 3.5 Plus 262K chat.qwen.ai
qwen-web/qwq QwQ 32K chat.qwen.ai
glm-web/glm-5 GLM-5 128K chatglm.cn
grok-web/grok-3 Grok 3 128K grok.com
gemini-web/gemini-3-flash Gemini 3 Flash 1M gemini.google.com
gemini-web/gemini-2.5-pro Gemini 2.5 Pro 1M gemini.google.com
perplexity-web/perplexity-default Perplexity 128K perplexity.ai
doubao-web/doubao-seed-2.0-pro Doubao Seed 2.0 Pro 256K doubao.com
xiaomimo-web/mimo-v2-pro MiMo V2 Pro 1M xiaomimimo.com

Configuration

Command Line Options

web-model-bridge                         # Start with defaults
web-model-bridge -p 8080                 # Custom port
web-model-bridge --host 0.0.0.0          # Allow remote access (use with --auth-token)
web-model-bridge --auth-token mysecret   # Require Bearer token
web-model-bridge --no-open               # Don't auto-open browser
web-model-bridge -v                      # Verbose logging (shows environment check)

Config File

~/.webmodel/config.yml:

server:
  port: 3456
  host: 127.0.0.1
  authToken: null

browser:
  idleShutdown: 300    # Close Chrome after 5 min idle

providers:
  enabled:             # Enable only what you need
    - claude-web
    - deepseek-web
    - qwen-web

logging:
  level: info

API Reference

OpenAI Format

curl http://localhost:3456/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-web/deepseek-v4",
    "messages": [{"role": "user", "content": "Hello"}],
    "stream": true
  }'

Anthropic Format

curl http://localhost:3456/v1/messages \
  -H "Content-Type: application/json" \
  -H "x-api-key: not-needed" \
  -H "anthropic-version: 2023-06-01" \
  -d '{
    "model": "claude-web/claude-sonnet-4-6",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Hello"}]
  }'

Management Endpoints

Endpoint Method Description
/ GET Web Dashboard
/v1/chat/completions POST OpenAI-compatible chat
/v1/messages POST Anthropic-compatible chat
/v1/models GET List available models
/webmodel/providers GET Provider auth status
/webmodel/health GET Server health check
/webmodel/auth/login POST Trigger provider login
/webmodel/auth/logout POST Clear provider auth

Architecture

┌──────────────────────────────┐
│  AI Tools (OpenClaw, Claude  │
│  Code, Cursor, Open WebUI)   │
└──────────┬───────────────────┘
           │ HTTP
           ▼
┌──────────────────────────────┐
│  web-model-bridge            │
│  ┌────────────────────────┐  │
│  │ HTTP Layer             │  │
│  │ OpenAI + Anthropic API │  │
│  └───────────┬────────────┘  │
│  ┌───────────▼────────────┐  │
│  │ Core Layer             │  │
│  │ Registry + SSE Stream  │  │
│  └───────────┬────────────┘  │
│  ┌───────────▼────────────┐  │
│  │ Infra Layer            │  │
│  │ Chrome + Auth + Config │  │
│  └────────────────────────┘  │
└──────────┬───────────────────┘
           │ CDP
           ▼
┌──────────────────────────────┐
│  Chrome (silent, background) │
│  Logged into AI websites     │
└──────────────────────────────┘

Troubleshooting

Problem Solution
"Chrome not found" Install Google Chrome. Run with -v to see detected paths
"Browser not connected" Chrome may have crashed. Restart web-model-bridge
Cookie expired Click Re-login in Dashboard — no restart needed
Port 3456 in use Use -p 8080 or any free port
Claude Code 404 ANTHROPIC_BASE_URL must NOT end with /v1
Cursor connection fails Some Cursor versions require ngrok for localhost
Windows: Chrome not detected Ensure Chrome is in default install path (Program Files)

Testing

npm test              # All tests (132 passing)
npm run test:unit     # Unit tests only
npm run test:integration  # Integration tests
npm run test:coverage # Coverage report
npm run typecheck     # TypeScript strict check

Development

git clone https://github.com/linuxhsj/WebModel.git
cd WebModel
npm install
npm run dev           # Start in dev mode
npm test              # Run tests
npm run build         # Build for distribution

Contributing

Contributions welcome! Areas where help is needed:

  • 🌐 New provider adapters
  • 🧪 E2E test coverage
  • 📱 Mobile-friendly Dashboard
  • 🐳 Docker image
  • 🔧 Real upstream API endpoint discovery

License

MIT