AnyProxyAi

Universal AI API Proxy Router with Multi-Format Support

English | 简体中文

💰 This tool is valued at $100 USD - developed with equivalent cost of Claude Opus 4.5 AI assistance

📖 Introduction

AnyProxyAi is a universal AI API proxy router that supports multiple API formats (OpenAI, Claude, Gemini) with automatic format conversion, load balancing, and intelligent routing. It provides a unified local endpoint for all your AI API needs.

Why AnyProxyAi?

• Multi-Format Support: Seamlessly convert between OpenAI, Claude (Anthropic), and Gemini API formats
• Unified Endpoint: Single local API endpoint for all AI services
• Load Balancing: Automatic routing across multiple API endpoints
• Usage Statistics: Real-time monitoring of requests, tokens, and success rates
• Cross-Platform: Native desktop app for Windows, macOS, and Linux
• Multi-Language: Supports Chinese and English, follows system language by default

📸 Screenshots

Home Page - Dashboard

Model List - Route Management

Usage Statistics - Heatmap & Charts

Add Route Dialog

Linux Running

Claude Code Interface → OpenAI Format Conversion

OpenAI Format Conversion

✨ Features

Core Features

Feature	Description
🔄 API Format Conversion	Automatic conversion between OpenAI, Claude, and Gemini formats
🔀 Smart Routing	Route requests to different backends based on model name
🔁 Proxy Redirect	Use proxy_auto keyword to redirect to any configured model
📊 Real-time Stats	Monitor requests, errors, and token usage
📈 Historical Data	SQLite-based statistics with heatmap visualization
🎯 Model Ranking	Track most used models and their performance
🌐 Multi-Language	Chinese/English support, follows system language, switchable from top-right corner (persistent)

Supported API Formats

Format	Input	Output	Streaming	Stream Conversion
OpenAI	✅	✅	✅	✅
Claude (Anthropic)	✅	✅	✅	✅
Gemini	✅	✅	✅	✅
Claude Code	✅	✅	✅	✅
Cursor IDE	✅	✅	✅	✅

UI Features

Feature	Description
🖥️ Cross-platform Desktop App	Windows, macOS, Linux support
🎨 Dark/Light Theme	Toggle between themes
🌐 Language Switch	Switch language from top-right corner popup (persistent)
📋 System Tray	Minimize to system tray
📝 Route Management	Add, edit, delete, import/export routes
📊 Usage Dashboard	Heatmap, charts, and statistics

🚀 Quick Start

Download

📥 Download Latest Release

Available Builds:

• anyproxyai-windows-amd64.exe - Windows x64
• anyproxyai-windows-arm64.exe - Windows ARM64
• anyproxyai-linux-amd64 - Linux x64
• anyproxyai-linux-arm64 - Linux ARM64
• anyproxyai-darwin-amd64.zip - macOS Intel
• anyproxyai-darwin-arm64.zip - macOS Apple Silicon

Windows

1. Download anyproxyai-windows-amd64.exe
2. Run the executable
3. Allow firewall access if prompted

macOS

1. Download anyproxyai-darwin-arm64.zip (Apple Silicon) or anyproxyai-darwin-amd64.zip (Intel)
2. Extract and move anyproxyai.app to Applications
3. First run: Right-click → Open (bypass Gatekeeper)

Linux

chmod +x anyproxyai-linux-amd64
./anyproxyai-linux-amd64

Setup

1. Add API Route

Click "添加路由" (Add Route) and configure:

Field	Description	Example
Name	Friendly name	GPT-4 Turbo
Model	Model identifier	gpt-4-turbo
API URL	Backend API URL	https://api.openai.com
API Key	Your API key	sk-xxx...
Group	Optional grouping	OpenAI
Format	API format type	openai / claude / gemini

2. Configure Your Application

Use the local proxy endpoint in your application:

OpenAI Compatible:

API Base URL: http://localhost:5642/api
API Key: (use the key shown on home page, or any value if auth is disabled)

Claude/Anthropic:

API Base URL: http://localhost:5642/api/anthropic
API Key: (use the key shown on home page, or any value if auth is disabled)

Claude Code:

API Base URL: http://localhost:5642/api/claudecode
API Key: (use the key shown on home page, or any value if auth is disabled)

Gemini:

API Base URL: http://localhost:5642/api/gemini
API Key: (use the key shown on home page, or any value if auth is disabled)

Cursor IDE:

API Base URL: http://localhost:5642/api/cursor/v1
API Key: (use the key shown on home page, or any value if auth is disabled)

Note about Cursor IDE: Cursor uses OpenAI-compatible endpoints but with Anthropic/Claude-style tool definitions and message formats. AnyProxyAi automatically detects and converts Cursor's format to standard OpenAI format, so you can use any OpenAI-compatible backend with Cursor.

Note: The API Key shown on the home page is used for authentication. If you want to disable authentication, set local_api_key to empty string in config.json.

3. Use Proxy Redirect (Optional)

Enable redirect and set proxy_auto as your model name to automatically route to your configured target model.

📖 Architecture

┌─────────────────┐     ┌─────────────────────────────────────────────────────┐
│  Your App       │────▶│                 AnyProxyAi                          │
│  (Any SDK)      │     │  localhost:5642                                     │
└─────────────────┘     │                                                     │
                        │  ┌─────────────────────────────────────────────┐   │
                        │  │              API Router                      │   │
                        │  │  /api/v1/*        → OpenAI format           │   │
                        │  │  /api/cursor/*    → Cursor IDE format       │   │
                        │  │  /api/anthropic/* → Claude format           │   │
                        │  │  /api/claudecode/*→ Claude Code format      │   │
                        │  │  /api/gemini/*    → Gemini format           │   │
                        │  └─────────────────────────────────────────────┘   │
                        │                        │                            │
                        │  ┌─────────────────────▼─────────────────────────┐ │
                        │  │           Format Adapters                      │ │
                        │  │  ┌──────────┐ ┌──────────┐ ┌──────────┐       │ │
                        │  │  │ OpenAI   │ │ Claude   │ │ Gemini   │       │ │
                        │  │  │ Adapter  │ │ Adapter  │ │ Adapter  │       │ │
                        │  │  └──────────┘ └──────────┘ └──────────┘       │ │
                        │  │  ┌──────────┐                                 │ │
                        │  │  │ Cursor   │ (Auto-detect & convert)         │ │
                        │  │  │ Adapter  │                                 │ │
                        │  │  └──────────┘                                 │ │
                        │  └───────────────────────────────────────────────┘ │
                        │                        │                            │
                        │  ┌─────────────────────▼─────────────────────────┐ │
                        │  │           Backend Routes (Cloud)               │ │
                        │  │  ┌──────────┐ ┌──────────┐ ┌──────────┐       │ │
                        │  │  │ OpenAI   │ │ Claude   │ │ Gemini   │       │ │
                        │  │  │ Cloud    │ │ Cloud    │ │ Cloud    │       │ │
                        │  │  └──────────┘ └──────────┘ └──────────┘       │ │
                        │  └───────────────────────────────────────────────┘ │
                        └─────────────────────────────────────────────────────┘

Request Flow

1. Receive Request: Local proxy receives API request
2. Route Matching: Find matching route by model name
3. Format Detection: Detect source format from request path
4. Adapter Selection: Choose appropriate adapter based on route format
5. Request Transformation: Convert request to target format
6. Backend Call: Forward to actual API endpoint
7. Response Transformation: Convert response back to source format
8. Statistics: Log request metrics

Adapter Matrix

Source → Target	OpenAI	Claude	Gemini
OpenAI	Pass-through	claude-to-openai	gemini-to-openai
Claude	openai-to-claude	Pass-through	gemini-to-claude
Gemini	openai-to-gemini	claude-to-gemini	Pass-through

🔧 Configuration

config.json

{
  "host": "localhost",
  "port": 5642,
  "database_path": "routes.db",
  "local_api_key": "sk-local-default-key",
  "redirect_enabled": true,
  "redirect_keyword": "proxy_auto",
  "redirect_target_model": "gpt-4-turbo",
  "minimize_to_tray": true,
  "auto_start": false
}

Route Configuration

Routes are stored in SQLite database (routes.db) with the following schema:

Field	Type	Description
name	TEXT	Display name
model	TEXT	Model identifier (used for routing)
api_url	TEXT	Backend API base URL
api_key	TEXT	API authentication key
group	TEXT	Optional grouping
format	TEXT	API format: openai, claude, gemini
enabled	INTEGER	1=enabled, 0=disabled

🛠️ Development

Requirements

• Go 1.22+
• Node.js 18+

Development Mode

# Install frontend dependencies
cd frontend && npm install && cd ..

# Run directly with Go
go run .

Build

# Build for current platform
go build -o anyproxyai .

# Or use the build script
./build.sh        # Linux/macOS
build.bat         # Windows

Project Structure

anyproxyai/
├── main.go                    # Application entry, Wails bindings
├── wails.json                 # Wails configuration
├── config.json                # Runtime configuration
│
├── internal/                  # Go backend modules
│   ├── adapters/              # API format converters
│   │   ├── adapter.go         # Adapter interface
│   │   ├── anthropic.go       # Claude adapter
│   │   ├── gemini.go          # Gemini adapter
│   │   ├── openai_to_claude.go
│   │   ├── claude_to_openai.go
│   │   └── ...
│   ├── config/                # Configuration management
│   ├── database/              # SQLite database
│   ├── router/                # HTTP router (Gin)
│   ├── service/               # Business logic
│   │   ├── proxy_service.go   # Proxy & streaming
│   │   └── route_service.go   # Route management
│   └── system/                # System tray, autostart
│
└── frontend/                  # Vue 3 frontend
    ├── src/
    │   ├── App.vue            # Main application
    │   ├── i18n/              # Internationalization
    │   └── components/        # UI components
    └── wailsjs/               # Wails bindings

🔨 GitHub Actions Build

This project uses GitHub Actions for automated multi-platform builds.

Trigger Build

Builds are triggered in the following ways:

Trigger	Build	Release	Example
package(...) commit	✅	❌	git commit -m "package(build): fix issue"
tag(vX.X.X): message commit	✅	✅	git commit -m "tag(v1.0.0): Initial release"
Push tag v*	✅	✅	git tag v1.0.0 && git push origin v1.0.0
Pull Request	✅	❌	PR to main/master
Manual trigger	✅	❌	workflow_dispatch

Quick Release (Recommended)

Use tag(version): description format in your commit message to automatically build, create tag, and publish release:

# This will: build all platforms → create tag v1.0.0 → publish release with description
git commit -m "tag(v1.0.0): Initial release with multi-format API support"
git push origin main

The release description will be automatically filled with the text after the colon.

Build Only (No Release)

Use package(...) prefix for build-only commits:

git commit -m "package(build): fix linux arm64 build"
git push origin main

Download Artifacts

1. Go to Actions tab in GitHub repository
2. Click on the completed workflow run
3. Download individual artifacts:
   • anyproxyai-windows-amd64
   • anyproxyai-windows-arm64
   • anyproxyai-linux-amd64
   • anyproxyai-linux-arm64
   • anyproxyai-darwin-amd64
   • anyproxyai-darwin-arm64

❓ FAQ

Installation

Q: Windows shows security warning? A: Click "More info" → "Run anyway". The app is not code-signed.

Q: macOS shows "cannot be opened" error? A: Right-click → Open → Open. Or allow in System Preferences → Security & Privacy.

Q: Port 5642 is already in use? A: Edit config.json and change the port value, or change it in Settings page.

Usage

Q: How does format conversion work? A: The proxy detects the incoming request format from the URL path and converts it to the target format based on the route's format setting.

Q: What is proxy_auto? A: A special keyword that redirects to your configured target model, allowing you to use a single model name across different applications.

Q: Are token counts accurate? A: Token counts are estimates based on response data. Actual billing may differ.

Q: How to switch language? A: Click the language icon in the top-right corner to open the language switch popup. The setting is persistent.

Development

Q: How to add a new adapter? A: Implement the Adapter interface in internal/adapters/ and register it in adapter.go.

🙏 References

This project was inspired by and references the following projects:

• ccNexus - Claude Code Nexus
• LLM-API-Transform-Proxy - LLM API Transform Proxy

📄 License

This project is open source under the MIT License.

---

Made with ❤️ and Claude AI From China