Skip to main content
Advanced Setup - This integration requires running a local proxy. For simpler setups, consider Aider or OpenCode.
Claude Code is Anthropic’s official CLI for AI-assisted coding. While it’s designed for Claude models, you can route requests to Infercom using ccproxy, a LiteLLM-based proxy.

​
How It Works

ccproxy intercepts Claude Code’s API calls and routes them to your configured provider (Infercom). This allows you to use Claude Code’s interface while running inference on EU sovereign infrastructure. 
Claude Code CLI -> ccproxy (localhost) -> Infercom API

​
Prerequisites

​
Installation

​
Step 1: Install ccproxy

# Using npm
npm install -g ccproxy

# Or using Bun
bun install -g ccproxy

​
Step 2: Create Configuration

Create ~/.ccproxy/config.yaml: 
model_list:
  # Route Sonnet requests to Infercom MiniMax
  - model_name: infercom
    litellm_params:
      model: openai/MiniMax-M2.5
      api_base: https://api.infercom.ai/v1
      api_key: "your-infercom-api-key"

  # Keep Claude models for complex tasks (optional)
  - model_name: claude-sonnet-4-6
    litellm_params:
      model: anthropic/claude-sonnet-4-6
      api_base: https://api.anthropic.com
      api_key: "your-anthropic-api-key"

litellm_settings:
  drop_params: true  # Required - drops unsupported params like reasoning_effort
Create ~/.ccproxy/ccproxy.yaml for routing rules: 
ccproxy:
  rules:
    # Route Sonnet to Infercom
    - name: infercom
      rule: ccproxy.rules.MatchModelRule
      params:
        - model_name: claude-sonnet-4-6

litellm:
  host: 127.0.0.1
  port: 4000
drop_params: true is required because MiniMax doesn’t support all Claude parameters like reasoning_effort.
Map multiple Claude model names to ensure compatibility with different Claude Code versions.

​
Step 3: Start ccproxy

cd ~/.ccproxy && ccproxy start
You should see: 
ccproxy running on http://127.0.0.1:4000
Keep this terminal open while using Claude Code.

​
Usage

​
Configure Claude Code

In a new terminal, set the API endpoint: 
export ANTHROPIC_BASE_URL="http://127.0.0.1:4000"

​
Run Claude Code

claude

​
Switch Models

Use the /model command in Claude Code to select a model. The model name will show as “Claude” but requests are routed to MiniMax-M2.5.

​
Running ccproxy as a Service

​
macOS (launchd)

Create ~/Library/LaunchAgents/com.ccproxy.plist: 
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
    <key>Label</key>
    <string>com.ccproxy</string>
    <key>ProgramArguments</key>
    <array>
        <string>/usr/local/bin/ccproxy</string>
    </array>
    <key>RunAtLoad</key>
    <true/>
    <key>KeepAlive</key>
    <true/>
</dict>
</plist>
Load the service: 
launchctl load ~/Library/LaunchAgents/com.ccproxy.plist

​
Linux (systemd)

Create ~/.config/systemd/user/ccproxy.service: 
[Unit]
Description=ccproxy for Claude Code
After=network.target

[Service]
ExecStart=/usr/local/bin/ccproxy
Restart=always

[Install]
WantedBy=default.target
Enable and start: 
systemctl --user enable ccproxy
systemctl --user start ccproxy

​
Limitations

Known Limitations:
  • Model name displays as “Claude” in the interface even when using Infercom
  • Some Claude-specific features may not work (vision, computer use)
  • Requires keeping proxy running
  • VS Code extension is less reliable than CLI

​
Troubleshooting

​
Connection Refused

Ensure ccproxy is running: 
curl http://127.0.0.1:4000/health

​
Authentication Errors

Verify your Infercom API key in ccproxy.yaml: 
curl -s https://api.infercom.ai/v1/models \
  -H "Authorization: Bearer your-infercom-api-key"

​
Model Not Found

Check that your ccproxy.yaml maps the correct Claude model names: 
model_name: "claude-sonnet-4-20250514"  # Must match what Claude Code requests

​
Slow Responses

ccproxy adds minimal latency. If responses are slow, check:
  1. Your network connection to api.infercom.ai
  2. Model load times (first request may be slower)

​
Security Considerations

  • ccproxy runs locally and doesn’t expose your API key to external services
  • All traffic to Infercom is encrypted (HTTPS)
  • Your code stays within the EU infrastructure

​
Alternative: Direct API

For simpler setups without the proxy overhead, consider:

​
Next Steps