Anthropic Provider

CCProxy provides native support for Anthropic's Claude models, including the latest Claude 4 family with enhanced reasoning and coding capabilities.

Latest Models (July 2025)

Claude 4 Family

claude-opus-4-20250720 - Most powerful model, best for complex reasoning and coding (SWE-bench 72.5%)
- Exceptional performance on complex tasks
- Superior reasoning and analysis capabilities
- Best choice for long-context processing (>60K tokens)
- Ideal for advanced coding tasks
claude-sonnet-4-20250720 - Balanced performance and speed
- Excellent balance of capabilities and response time
- Great for general-purpose tasks
- Strong reasoning with faster responses than Opus
- Recommended as default model

Claude 3 Family

claude-3-7-sonnet-20250721 - Most intelligent Claude 3 model
- Latest iteration of Claude 3 Sonnet
- Enhanced capabilities over previous Claude 3 models
- Good alternative when Claude 4 models are not required
claude-3-5-haiku-20241022 - Fast and efficient for simple tasks
- Optimized for speed and cost-effectiveness
- Ideal for background tasks and simple queries
- Low latency responses

Configuration

Provider Setup

json

{
  "providers": {
    "anthropic": {
      "enabled": true,
      "api_key": "${ANTHROPIC_API_KEY}",
      "base_url": "https://api.anthropic.com/v1",
      "models": {
        "claude-opus-4-20250720": {
          "enabled": true,
          "max_tokens": 200000,
          "context_window": 200000
        },
        "claude-sonnet-4-20250720": {
          "enabled": true,
          "max_tokens": 200000,
          "context_window": 200000
        },
        "claude-3-7-sonnet-20250721": {
          "enabled": true,
          "max_tokens": 200000,
          "context_window": 200000
        },
        "claude-3-5-haiku-20241022": {
          "enabled": true,
          "max_tokens": 200000,
          "context_window": 200000
        }
      }
    }
  }
}

Routing Configuration

json

{
  "router": {
    "routes": {
      "default": {
        "provider": "anthropic",
        "model": "claude-sonnet-4-20250720",
        "description": "Balanced performance for most tasks"
      },
      "longContext": {
        "provider": "anthropic",
        "model": "claude-opus-4-20250720",
        "description": "Best for >60K tokens or complex reasoning",
        "token_threshold": 60000
      },
      "background": {
        "provider": "anthropic",
        "model": "claude-3-5-haiku-20241022",
        "description": "Fast, cost-effective for simple tasks"
      }
    }
  }
}

Environment Variables

bash

# Required - CCProxy will auto-detect this
export ANTHROPIC_API_KEY="sk-ant-..."

# Optional - for CCProxy to intercept Claude Code
export ANTHROPIC_BASE_URL="http://127.0.0.1:3456"

Request Format

CCProxy accepts standard Anthropic API format with the following supported parameters:

json

{
  "model": "claude-sonnet-4-20250720",
  "messages": [
    {
      "role": "user",
      "content": "Hello, Claude!"
    }
  ],
  "max_tokens": 1024,
  "temperature": 0.7,
  "top_p": 0.9
}

Supported Parameters

model (required): The model identifier
messages (required): Array of message objects with role and content
max_tokens (optional): Maximum tokens to generate (default: model-specific)
temperature (optional): Sampling temperature between 0 and 1 (default: 1.0)
top_p (optional): Nucleus sampling between 0 and 1 (default: 1.0)
stream (optional): Enable streaming responses (default: false)
system (optional): System message (extracted from messages array)
tools (optional): Array of tool definitions for function calling
tool_choice (optional): Tool selection strategy

Removed Parameters

The following OpenAI-style parameters are automatically removed when routing to Anthropic:

frequency_penalty: Not supported by Anthropic API
presence_penalty: Not supported by Anthropic API

These parameters will be silently dropped to ensure compatibility.

Streaming Support

All Anthropic models support streaming responses:

bash

curl -X POST http://localhost:3456/v1/messages \
  -H "Content-Type: application/json" \
  -H "x-api-key: your-api-key" \
  -H "anthropic-version: 2023-06-01" \
  -d '{
    "model": "claude-sonnet-4-20250720",
    "messages": [{"role": "user", "content": "Stream a response"}],
    "stream": true,
    "max_tokens": 1000
  }'

Model Selection Guidelines

Use claude-opus-4-20250720 for:

Complex reasoning tasks requiring deep analysis
Long-context processing (>60K tokens)
Advanced coding and software engineering tasks
Tasks requiring maximum intelligence and capability

Use claude-sonnet-4-20250720 for:

General-purpose conversations
Balanced performance requirements
Standard coding tasks
Most default routing scenarios

Use claude-3-7-sonnet-20250721 for:

When Claude 4 models are unavailable
Tasks requiring Claude 3 specific behavior
Compatibility with older integrations

Use claude-3-5-haiku-20241022 for:

High-volume, simple tasks
Background processing
Cost-sensitive applications
Low-latency requirements

Advanced Features

Context Window Management

All models support 200K token context windows. CCProxy automatically handles:

Token counting using tiktoken-compatible tokenizer
Automatic routing to longContext models when needed
Context window overflow prevention

Tool Use

All Anthropic models support function calling:

json

{
  "model": "claude-sonnet-4-20250720",
  "messages": [
    {
      "role": "user",
      "content": "What's the weather like?"
    }
  ],
  "tools": [
    {
      "name": "get_weather",
      "description": "Get current weather",
      "input_schema": {
        "type": "object",
        "properties": {
          "location": {"type": "string"}
        }
      }
    }
  ]
}

Error Handling

CCProxy provides standardized error responses:

json

{
  "error": {
    "type": "invalid_request_error",
    "message": "Model not found: claude-3-opus-20240229",
    "suggestion": "Use claude-opus-4-20250720 instead"
  }
}

Rate Limiting

Anthropic enforces rate limits per model tier:

Opus models: Lower rate limits, higher priority
Sonnet models: Balanced rate limits
Haiku models: Higher rate limits for volume usage

CCProxy respects these limits and provides appropriate retry mechanisms.

Best Practices

Model Selection: Choose the appropriate model based on task complexity
Token Management: Monitor token usage to optimize routing
Streaming: Use streaming for better user experience on long responses
Error Handling: Implement retry logic for transient errors
Cost Optimization: Use Haiku for simple tasks to reduce costs
Context Management: Be mindful of context window limits

Migration from Older Models

If you're using deprecated models, update your configuration:

Old Model	Recommended Replacement
claude-3-opus-20240229	claude-opus-4-20250720
claude-3-sonnet-20240229	claude-sonnet-4-20250720
claude-3-haiku-20240307	claude-3-5-haiku-20241022
claude-2.1	claude-sonnet-4-20250720
claude-2.0	claude-sonnet-4-20250720
claude-instant-1.2	claude-3-5-haiku-20241022

Anthropic Provider ​

Latest Models (July 2025) ​

Claude 4 Family ​

Claude 3 Family ​

Configuration ​

Provider Setup ​

Routing Configuration ​

Environment Variables ​

Request Format ​

Supported Parameters ​

Removed Parameters ​

Streaming Support ​

Model Selection Guidelines ​

Use claude-opus-4-20250720 for: ​

Use claude-sonnet-4-20250720 for: ​

Use claude-3-7-sonnet-20250721 for: ​

Use claude-3-5-haiku-20241022 for: ​

Advanced Features ​

Context Window Management ​

Tool Use ​

Error Handling ​

Rate Limiting ​

Best Practices ​

Migration from Older Models ​

Anthropic Provider

Latest Models (July 2025)

Claude 4 Family

Claude 3 Family

Configuration

Provider Setup

Routing Configuration

Environment Variables

Request Format

Supported Parameters

Removed Parameters

Streaming Support

Model Selection Guidelines

Use claude-opus-4-20250720 for:

Use claude-sonnet-4-20250720 for:

Use claude-3-7-sonnet-20250721 for:

Use claude-3-5-haiku-20241022 for:

Advanced Features

Context Window Management

Tool Use

Error Handling

Rate Limiting

Best Practices

Migration from Older Models