High-Performance LLM Proxy for the Modern Era

🌍 Language Support

🇺🇸 English | 🇨🇳 简体中文

---

🎯 What is ClewdR?

ClewdR is a production-grade, high-performance proxy server engineered specifically for Claude (Claude.ai, Claude Code) and Google Gemini (AI Studio, Vertex AI). Built with Rust for maximum performance and minimal resource usage, it provides enterprise-level reliability with consumer-friendly simplicity.

🏆 Why ClewdR?

• 🚄 10x Performance: Outperforms script-language implementations
• 💾 1/10th Memory: Uses only single-digit MB in production
• 🔧 Production Ready: Handles thousands of requests per second
• 🌐 Multi-Platform: Native support for Windows, macOS, Linux, Android

✨ Core Features

🎨 Full-Featured Web Interface React-powered dashboard with real-time monitoring Multi-language support (English/Chinese) Secure authentication with auto-generated passwords Hot configuration reload without service interruption Visual cookie & key management 🏗️ Enterprise Architecture Tokio + Axum async runtime for maximum throughput Event-driven design with decoupled components Moka-powered caching with intelligent invalidation Chrome-level fingerprinting for seamless API access Multi-threaded processing with optimal resource usage 🧠 Intelligent Resource Management Smart cookie rotation with status classification API key health monitoring and automatic failover Rate limiting protection with exponential backoff Connection pooling with keep-alive optimization

🌍 Universal Compatibility Static compilation - single binary, zero dependencies Cross-platform native - Windows, macOS, Linux, Android Docker ready with optimized images Reverse proxy friendly with custom endpoint support 🚀 Protocol Support Claude Integration ✅ Claude.ai web interface ✅ Claude Code specialized support ✅ System prompt caching for efficiency ✅ Extended Thinking mode ✅ Image attachments & web search ✅ Custom stop sequences Google Gemini Integration ✅ AI Studio & Vertex AI ✅ OAuth2 authentication for Vertex ✅ HTTP Keep-Alive optimization ✅ Model switching with automatic detection API Compatibility ✅ OpenAI format - drop-in replacement ✅ Native formats - Claude & Gemini ✅ Streaming responses with real-time processing

📊 Performance Metrics

Metric	ClewdR	Traditional Proxies
Memory Usage	<10 MB	100-500 MB
Requests/sec	1000+	100-200
Startup Time	<1 second	5-15 seconds
Binary Size	~15 MB	50-200 MB
Dependencies	Zero	Node.js/Python + libs

🚀 Quick Start Guide

Step 1: Download & Run

# Download the latest release for your platform
wget https://github.com/Xerxes-2/clewdr/releases/latest/download/clewdr-[platform]

# Extract the binary (if necessary)
tar -xzf clewdr-[platform].tar.gz

# Navigate to the directory
cd clewdr-[platform]

# Make executable (Linux/macOS)
chmod +x clewdr

# Run ClewdR
./clewdr

📦 Platform Downloads

Platform	Architecture	Download Link
🪟 Windows	x64	clewdr-windows-x64.exe
🐧 Linux	x64	clewdr-linux-x64
🐧 Linux	ARM64	clewdr-linux-arm64
🍎 macOS	x64	clewdr-macos-x64
🍎 macOS	ARM64 (M1/M2)	clewdr-macos-arm64
🤖 Android	ARM64	clewdr-android-arm64

Step 2: Access Web Interface

1. 🌐 Open your browser to http://127.0.0.1:8484
2. 🔐 Use the Web Admin Password displayed in the console
3. 🎉 Welcome to ClewdR's management interface!

💡 Pro Tips:

• Forgot password? Delete clewdr.toml and restart
• Docker users: Password appears in container logs
• Change password: Use the web interface settings

Step 3: Configure Your Services

🍃 Claude Setup Add Cookies: Paste your Claude.ai session cookies Configure Proxy: Set upstream proxy if needed Test Connection: Verify cookie status in dashboard

🔹 Gemini Setup Add API Keys: Input your Google AI Studio keys Vertex AI (Optional): Configure OAuth2 for enterprise Model Selection: Choose your preferred models

Step 4: Connect Your Applications

ClewdR provides multiple API endpoints. Check the console output for available endpoints:

🔗 API Endpoints

# Claude Endpoints
Claude Web:    http://127.0.0.1:8484/v1/messages          # Native format
Claude OpenAI: http://127.0.0.1:8484/v1/chat/completions  # OpenAI compatible
Claude Code:   http://127.0.0.1:8484/code/v1/messages     # Claude Code

# Gemini Endpoints  
Gemini Native: http://127.0.0.1:8484/v1/v1beta/generateContent    # Native format
Gemini OpenAI: http://127.0.0.1:8484/gemini/chat/completions      # OpenAI compatible
Vertex AI:     http://127.0.0.1:8484/v1/vertex/v1beta/            # Vertex AI

⚙️ Application Configuration Examples

SillyTavern Configuration

{
  "api_url": "http://127.0.0.1:8484/v1/chat/completions",
  "api_key": "your-api-password-from-console",
  "model": "claude-3-sonnet-20240229"
}

Continue VSCode Extension

{
  "models": [
    {
      "title": "Claude via ClewdR",
      "provider": "openai",
      "model": "claude-3-sonnet-20240229",
      "apiBase": "http://127.0.0.1:8484/v1/",
      "apiKey": "your-api-password-from-console"
    }
  ]
}

Cursor IDE Configuration

{
  "openaiApiBase": "http://127.0.0.1:8484/v1/",
  "openaiApiKey": "your-api-password-from-console"
}

Step 5: Verify & Monitor

• ✅ Check cookie/key status in the web dashboard
• ✅ Monitor request logs for successful connections
• ✅ Test with a simple chat request
• ✅ Enjoy blazing-fast LLM proxy performance!

🗃️ Database Persistence

ClewdR keeps state in a local clewdr.toml file by default. To persist configuration, cookies, and API keys in a database instead, compile the binary with the database feature set and point persistence.mode at your driver.

Enable database features

• Build from source with the matching feature flag: cargo build --release --no-default-features --features "embed-resource,xdg,db-sqlite"
• Choose db-sqlite, db-postgres, or db-mysql (they automatically include the base db feature)
• Custom Docker images should adjust the cargo build step to include the desired db-* feature; the published Dockerfile uses file mode by default

Configure the connection

Add a persistence section to clewdr.toml (or set the equivalent environment variables):

[persistence]
mode = "postgres"            # sqlite | postgres | mysql
database_url = "postgres://user:pass@db:5432/clewdr"

• SQLite: optionally set sqlite_path = "/var/lib/clewdr/clewdr.db"; ClewdR expands this to sqlite:///var/lib/clewdr/clewdr.db?mode=rwc and creates the parent folder when possible

• Postgres/MySQL: database_url must be provided (for example postgres://user:pass@host:5432/db)

• Environment variable form uses Figment’s double-underscore syntax, e.g.:

  export CLEWDR_PERSISTENCE__MODE=sqlite
  export CLEWDR_PERSISTENCE__SQLITE_PATH=/var/lib/clewdr/clewdr.db
  # or database_url for server backends
  export CLEWDR_PERSISTENCE__DATABASE_URL="postgres://user:pass@db/clewdr"

Operational notes & precautions

• On first start ClewdR runs automatic SeaORM migrations (config, cookies, keys, wasted tables); ensure the account can create tables and indexes
• API endpoints that write cookies/keys check GET /api/storage/status; failed connections surface as Database storage is unavailable
• SQLite paths should live on persistent storage (bind mount the directory when running in containers)
• Setting persistence.mode without a matching db-* build leaves ClewdR in file mode—verify your binary with clewdr -V or rebuild with the correct features
• The admin API exposes helpers: GET /api/storage/status to inspect health, and authenticated POST /api/storage/import|export for file migration

Community Resources

Github Aggregated Wiki: https://github.com/Xerxes-2/clewdr/wiki

• Database persistence guide (中文)

Acknowledgements

• wreq - Excellent browser fingerprinting library used for API access.
• Clewd Modified Version - A modified version of the original Clewd, providing many inspirations and foundational features.
• Clove - Provides the support logic for Claude Code.