CCProxy

Appearance

OpenRouter Provider ​

OpenRouter unlocks the full AI ecosystem by providing access to 100+ different AI models through a single API, including the top-performing Qwen3 235B with its record-breaking AIME25 score of 70.3. As a major supporter of cutting-edge models, OpenRouter offers reliable access to Qwen3 235B, Kimi K2's ultra-fast inference, Grok's real-time data capabilities, alongside Claude, GPT-4, and many others. When combined with Claude Code and CCProxy, OpenRouter becomes the ultimate platform for AI experimentation, allowing you to leverage the most advanced models for each development task.

🎯 Why Choose OpenRouter for Claude Code? ​

  • 🏆 Top performance: Access to Qwen3 235B with AIME25 score of 70.3 (vs GPT-4o's 26.7)
  • ⚡ Ultra-fast options: Kimi K2 with 128K context for lightning-fast inference
  • 🌐 Real-time data: Grok models with current web information access
  • 🎯 100+ models: Access to all major AI models through familiar Claude Code interface
  • 🔄 Model fallbacks: Automatic failover between models for maximum reliability
  • 📊 Usage analytics: Detailed tracking and monitoring integrated with CCProxy
  • 💰 Cost optimization: Compare and choose the most cost-effective models for each task
  • 🧪 Perfect for experimentation: Test different models without changing your Claude Code workflow
  • ⚡ Unified API: Single integration for Qwen3, Kimi K2, Grok, Claude 3.5, GPT-4, and more

Setup ​

1. Get an API Key ​

  1. Visit openrouter.ai
  2. Sign up for an account
  3. Go to the API Keys section
  4. Generate a new API key

2. Configure CCProxy ​

Set the following environment variables:

bash
export PROVIDER=openrouter
export OPENROUTER_API_KEY=sk-or-v1-your_openrouter_api_key_here

3. Optional Configuration ​

bash
# Custom model (default: anthropic/claude-3.5-sonnet)
export OPENROUTER_MODEL=moonshotai/kimi-k2-instruct

# Custom max tokens (default: 16384)
export OPENROUTER_MAX_TOKENS=8192

# Custom base URL (default: https://openrouter.ai/api/v1)
export OPENROUTER_BASE_URL=https://openrouter.ai/api/v1

# Site URL for tracking (optional)
export OPENROUTER_SITE_URL=https://yourapp.com

# Site name for tracking (optional)
export OPENROUTER_SITE_NAME="Your App Name"

🌟 Top Models Available ​

Qwen3 235B A22B 2507 - Leading Performance ​

The current top-performing model on OpenRouter, Qwen3 235B delivers exceptional capabilities:

  • 🏆 AIME25 Score: 70.3 - Crushing GPT-4o's 26.7 score
  • 💡 Advanced reasoning: State-of-the-art mathematical and logical capabilities
  • 📊 Massive scale: 235B parameters with optimized inference
  • 🚀 Production ready: Reliable performance for demanding applications

Kimi K2 - Ultra-Fast Inference ​

  • ⚡ Lightning fast: Optimized for rapid responses
  • 📄 128K context: Handle large documents and codebases
  • 🎯 Tool calling: Full Claude Code compatibility
  • 💰 Cost effective: Great performance-to-price ratio

Grok Models - Real-Time Data ​

  • 🌐 Real-time access: Current information and web data
  • 🔄 Dynamic updates: Always up-to-date responses
  • 🛠️ Tool support: Compatible with Claude Code workflows
  • 📈 Continuous learning: Incorporates latest information

Available Models ​

OpenRouter provides access to 100+ AI models from leading providers including:

  • Qwen - Top-performing Qwen3 235B with record-breaking benchmarks
  • Anthropic - Claude series with advanced reasoning
  • OpenAI - GPT-4 series and reasoning models
  • Moonshot AI - Kimi K2 with ultra-fast inference
  • xAI - Grok models with real-time data access
  • Meta - Llama models for open-source applications
  • Google - Gemini models with multimodal capabilities
  • Mistral - European privacy-focused models
  • Many others - Including specialized coding and reasoning models

🔧 Critical for Claude Code: You must select models that support tool calling or function calling capabilities, as Claude Code requires these features to operate correctly.

Model Selection Guidelines ​

When choosing models on OpenRouter:

  1. Verify Tool Support: Ensure the model supports function calling
  2. Check Availability: Model availability changes frequently
  3. Review Capabilities: Different models excel at different tasks
  4. Consider Cost: Pricing varies significantly between models

For current model availability, capabilities, and pricing, visit OpenRouter's official models page.

Pricing ​

OpenRouter offers competitive pricing with transparent costs:

Free Tier ​

  • Free credits for new users
  • No monthly fees
  • Pay-as-you-use pricing

Cost Structure ​

  • Competitive per-token pricing across all models
  • No minimum spend required
  • Volume discounts available

For current, accurate pricing information, visit OpenRouter's official pricing page.

Configuration Examples ​

Basic Setup ​

bash
# .env file
PROVIDER=openrouter
OPENROUTER_API_KEY=sk-or-v1-your_api_key_here

Top Performance Setup - Qwen3 235B ​

bash
# For the highest performing model (AIME25: 70.3)
PROVIDER=openrouter
OPENROUTER_API_KEY=sk-or-v1-your_api_key_here
OPENROUTER_MODEL=qwen/qwen-3-235b-a22b-2507
OPENROUTER_MAX_TOKENS=16384

Ultra-Fast Setup - Kimi K2 ​

bash
# For ultra-fast inference with large context
PROVIDER=openrouter
OPENROUTER_API_KEY=sk-or-v1-your_api_key_here
OPENROUTER_MODEL=moonshotai/kimi-k2-instruct
OPENROUTER_MAX_TOKENS=8192

Real-Time Data Setup - Grok ​

bash
# For real-time information access
PROVIDER=openrouter
OPENROUTER_API_KEY=sk-or-v1-your_api_key_here
OPENROUTER_MODEL=xai/grok-beta
OPENROUTER_MAX_TOKENS=8192

High-Performance Setup ​

bash
# For speed-focused applications
PROVIDER=openrouter
OPENROUTER_API_KEY=sk-or-v1-your_api_key_here
OPENROUTER_MODEL=openai/gpt-4o-mini
OPENROUTER_MAX_TOKENS=4096

Quality-Focused Setup ​

bash
# For best quality responses
PROVIDER=openrouter
OPENROUTER_API_KEY=sk-or-v1-your_api_key_here
OPENROUTER_MODEL=anthropic/claude-3-opus
OPENROUTER_MAX_TOKENS=16384

Cost-Optimized Setup ​

bash
# For cost-effective usage
PROVIDER=openrouter
OPENROUTER_API_KEY=sk-or-v1-your_api_key_here
OPENROUTER_MODEL=qwen/qwen-2.5-72b-instruct
OPENROUTER_MAX_TOKENS=8192

Usage with Claude Code ​

Once configured, use Claude Code normally:

bash
# Set CCProxy as the API endpoint
export ANTHROPIC_BASE_URL=http://localhost:3456
# Claude Code will use CCProxy, no direct Anthropic API key needed

# Use Claude Code
claude "Compare different sorting algorithms"

Features ​

✅ Supported ​

  • Text generation
  • Function calling
  • Tool use
  • Streaming responses
  • Vision capabilities (model dependent)
  • JSON mode (model dependent)
  • Custom temperature/top_p
  • Model fallbacks
  • Usage tracking

⚠️ Model Dependent ​

  • Vision/image input
  • Real-time data access
  • Code execution
  • File uploads

Advanced Features ​

Model Routing Examples ​

Configure intelligent model routing based on task requirements:

bash
# For maximum performance tasks
export OPENROUTER_MODEL=qwen/qwen-3-235b-a22b-2507
export OPENROUTER_FALLBACK=anthropic/claude-3-opus

# For speed-critical applications  
export OPENROUTER_MODEL=moonshotai/kimi-k2-instruct
export OPENROUTER_FALLBACK=openai/gpt-4o-mini

# For real-time information needs
export OPENROUTER_MODEL=xai/grok-beta
export OPENROUTER_FALLBACK=xai/grok-2

Model Fallbacks ​

Configure automatic failover between models:

bash
# Set primary and fallback models
export OPENROUTER_MODEL=anthropic/claude-3.5-sonnet
export OPENROUTER_FALLBACK=openai/gpt-4o

Usage Tracking ​

OpenRouter provides detailed analytics:

bash
# Add tracking headers
export OPENROUTER_SITE_URL=https://yourapp.com
export OPENROUTER_SITE_NAME="Your App Name"

Custom Headers ​

bash
# Add custom tracking
export OPENROUTER_X_TITLE="Your Request Title"

Performance Tips ​

1. Choose the Right Model ​

bash
# For maximum performance: Use the top-scoring model
export OPENROUTER_MODEL=qwen/qwen-3-235b-a22b-2507

# For speed: Use ultra-fast models
export OPENROUTER_MODEL=moonshotai/kimi-k2-instruct

# For quality: Use larger, more capable models
export OPENROUTER_MODEL=anthropic/claude-3-opus

# For real-time data: Use models with web access
export OPENROUTER_MODEL=xai/grok-beta

# For cost: Use open-source models
export OPENROUTER_MODEL=qwen/qwen-2.5-72b-instruct

2. Optimize Token Usage ​

bash
# Reduce max tokens for faster responses
export OPENROUTER_MAX_TOKENS=1024

# Use appropriate context length
export OPENROUTER_MAX_TOKENS=4096

3. Monitor Usage ​

Check your usage and costs:

bash
# View OpenRouter dashboard
curl -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  https://openrouter.ai/api/v1/auth/key

Troubleshooting ​

Rate Limit Errors ​

json
{
  "error": {
    "message": "Rate limit exceeded",
    "type": "rate_limit_error"
  }
}

Solution: OpenRouter has generous rate limits. Wait and retry, or check your usage.

Authentication Errors ​

json
{
  "error": {
    "message": "Invalid API key",
    "type": "authentication_error"
  }
}

Solution: Verify your API key is correct and has sufficient credits.

Model Not Available ​

json
{
  "error": {
    "message": "Model not found or not available",
    "type": "invalid_request_error"
  }
}

Solution: Check the OpenRouter models page for available models.

Insufficient Credits ​

json
{
  "error": {
    "message": "Insufficient credits",
    "type": "insufficient_quota"
  }
}

Solution: Add credits to your OpenRouter account.

Model Selection Guide ​

For Top Performance ​

  • qwen/qwen-3-235b-a22b-2507 - #1 Overall - AIME25 score 70.3 (vs GPT-4o's 26.7)
  • anthropic/claude-3-opus - Highest quality responses
  • openai/o1-preview - Advanced reasoning

For Ultra-Fast Speed ​

  • moonshotai/kimi-k2-instruct - Fastest inference with 128K context
  • openai/gpt-4o-mini - Fast and cost-effective
  • anthropic/claude-3-haiku - Ultra-fast responses

For Real-Time Data ​

  • xai/grok-beta - Real-time web access and current information
  • xai/grok-2 - Enhanced reasoning with real-time data

For General Use ​

  • anthropic/claude-3.5-sonnet - Best all-around performance
  • openai/gpt-4o - Strong reasoning and tool use

For Cost ​

  • qwen/qwen-2.5-72b-instruct - Great quality-to-cost ratio
  • meta-llama/llama-3.1-405b-instruct - Open source powerhouse

For Coding ​

  • deepseek/deepseek-coder - Specialized for code
  • anthropic/claude-3.5-sonnet - Excellent code understanding
  • qwen/qwen-3-235b-a22b-2507 - Superior problem solving

Monitoring ​

Monitor your OpenRouter usage:

bash
# Check CCProxy logs
tail -f ccproxy.log

# Check OpenRouter status
curl http://localhost:3456/status

# View usage analytics on OpenRouter dashboard

Next Steps ​

  • Explore other providers for comparison and specialized use cases
  • Learn about model fallbacks for production reliability
  • Set up usage monitoring to optimize model selection and costs
  • Try Groq with Kimi K2 for ultra-fast inference alongside OpenRouter