CCLM - Claude Code with LM Studio

Use the official Claude Code CLI with local LM Studio models from anywhere in your terminal

Features

✅ Official Claude Code Interface - Full compatibility with Claude Code CLI ✅ Local LM Studio Models - Use any model loaded in LM Studio ✅ Global Command - Run cclm from any directory ✅ Auto Proxy Management - Automatic start/stop of translation proxy ✅ Zero Configuration - Works out of the box with sensible defaults ✅ Configurable - Easy configuration via CLI commands

Prerequisites

1. LM Studio - Download from https://lmstudio.ai/
2. Node.js 18+ - Download from https://nodejs.org/

Installation

Global Installation (Recommended)

npm install -g cclm

Local Installation

git clone https://github.com/yourusername/cclm
cd cclm
npm install
npm link

Quick Start

1. Start LM Studio

   • Open LM Studio
   • Load a model (e.g., Granite 4.0 H 1B, Llama 3.1 8B)
   • Go to "Local Server" tab
   • Click "Start Server"

2. Run CCLM

   cclm

That's it! You're now using Claude Code with your local LM Studio model.

Usage

Basic Commands

# Start interactive session
cclm

# Show help
cclm --help

# Show version
cclm --version

# Pass arguments to Claude Code
cclm --model my-model

Configuration

# Show current configuration
cclm config show

# Set LM Studio model
cclm config set lmStudioModel granite-4.0-h-1b

# Set LM Studio URL (if using custom port)
cclm config set lmStudioUrl http://localhost:1234/v1

# Enable debug mode
cclm config set debug true

# Change proxy port
cclm config set proxyPort 3001

# Disable auto-start proxy
cclm config set autoStartProxy false

Configuration Options

| Key | Default | Description | |-----|---------|-------------| | lmStudioUrl | http://localhost:1234/v1 | LM Studio API URL | | lmStudioModel | local-model | Model name in LM Studio | | proxyPort | 3000 | Port for proxy server | | debug | false | Enable debug logging | | autoStartProxy | true | Auto-start proxy if not running |

How It Works

┌─────────────┐       ┌──────────┐       ┌───────────┐
│  cclm CLI   │ ─────>│  Proxy   │ ─────>│ LM Studio │
│  (Official) │ API   │  Server  │ API   │  Server   │
└─────────────┘       └──────────┘       └───────────┘
   Anthropic           Translator        OpenAI format
   format

CCLM consists of three parts:

1. CCLM Wrapper (cclm command) - Manages setup and launches components
2. Proxy Server - Translates between Anthropic API and LM Studio (OpenAI) API
3. Claude Code CLI - Official Anthropic CLI (installed as dependency)

Features in Detail

Full Claude Code Compatibility

All official Claude Code features work:

• ✅ File operations (Read, Write, Edit)
• ✅ Search (Glob, Grep)
• ✅ Terminal commands (Bash)
• ✅ Task management
• ✅ Agents and plugins
• ✅ Streaming responses
• ✅ Function/tool calling
• ✅ MCP servers
• ✅ Slash commands

Auto Proxy Management

The proxy server starts automatically when needed:

# First run - proxy starts automatically
$ cclm

[1/3] Checking LM Studio... ✓
[2/3] Checking proxy server... starting...
      ✓ Proxy server started
[3/3] Starting Claude Code... ✓

On subsequent runs, if the proxy is still running:

$ cclm

[1/3] Checking LM Studio... ✓
[2/3] Checking proxy server... ✓
[3/3] Starting Claude Code... ✓

Configuration Persistence

Your configuration is saved to config.json in the package directory and persists across updates.

Examples

Use Specific Model

# Configure once
cclm config set lmStudioModel llama-3.1-70b

# Then just run
cclm

Debug Mode

# Enable debug logging
cclm config set debug true

# Run with debug output
cclm

Custom Port

# If LM Studio uses a different port
cclm config set lmStudioUrl http://localhost:8080/v1

# Or if you want proxy on different port
cclm config set proxyPort 8000

Working Directory

CCLM works from any directory:

# Navigate to your project
cd ~/projects/my-app

# Run cclm
cclm

# Claude Code operates in current directory

Troubleshooting

LM Studio Not Detected

Error: LM Studio is not running or not accessible

Solution:

1. Verify LM Studio is running
2. Check that local server is enabled
3. Test manually: curl http://localhost:1234/v1/models
4. If using different port: cclm config set lmStudioUrl http://localhost:YOUR_PORT/v1

Proxy Won't Start

Error: Failed to start proxy server

Solution:

1. Check if port 3000 is available: netstat -ano | findstr :3000 (Windows) or lsof -ti:3000 (Mac/Linux)
2. Use different port: cclm config set proxyPort 3001
3. Enable debug: cclm config set debug true and check logs

Command Not Found

Error: cclm: command not found

Solution:

# Verify installation
npm list -g cclm

# Reinstall globally
npm install -g cclm

# Or link locally
cd cclm
npm link

Model Not Responding

Solution:

1. Verify model name matches LM Studio: cclm config set lmStudioModel your-exact-model-name
2. Check LM Studio logs for errors
3. Try smaller model or longer timeout
4. Enable debug mode to see requests: cclm config set debug true

Advanced Usage

Use with Different LLM Providers

CCLM works with any OpenAI-compatible API:

# Ollama
cclm config set lmStudioUrl http://localhost:11434/v1

# LocalAI
cclm config set lmStudioUrl http://localhost:8080/v1

# Text Generation WebUI
cclm config set lmStudioUrl http://localhost:5000/v1

Programmatic Usage

import { spawn } from 'child_process';

const cclm = spawn('cclm', ['--model', 'my-model'], {
  env: process.env,
  stdio: 'inherit'
});

Multiple Configurations

# Save current config
cp ~/.cclm/config.json ~/.cclm/config-granite.json

# Switch configs
cp ~/.cclm/config-llama.json ~/.cclm/config.json
cclm

Comparison with Alternatives

| Feature | CCLM | Custom CLI | API Direct | |---------|------|------------|------------| | Official CLI | ✅ | ❌ | ❌ | | Full Features | ✅ | ⚠️ Limited | ❌ | | Auto Updates | ✅ | ❌ | ❌ | | Easy Setup | ✅ | ⚠️ Medium | ❌ | | LM Studio | ✅ | ✅ | ✅ | | Global Command | ✅ | ⚠️ Manual | N/A |

Performance

Typical performance with Granite 4.0 H 1B (1M context):

• Startup: ~3-5 seconds
• First response: ~2-5 seconds
• Streaming: Real-time, word-by-word
• Tool execution: <100ms proxy overhead
• Memory: ~50MB (proxy) + ~2GB (LM Studio model)

Security & Privacy

• ✅ Runs 100% locally - no data leaves your machine
• ✅ No telemetry or tracking
• ✅ Proxy only accessible on localhost
• ✅ Dummy API key (not used externally)

Development

Build from Source

git clone <repository>
cd cclm-package
npm install
npm link

Run Tests

npm test

Debug

# Enable debug mode
cclm config set debug true

# Or set environment variable
DEBUG=true cclm

Changelog

1.0.0 (Initial Release)

• Official Claude Code CLI integration
• LM Studio proxy server
• Auto proxy management
• Configuration system
• Global command installation

License

MIT

Credits

• Claude Code - Official CLI by Anthropic
• LM Studio - Local LLM hosting platform
• CCLM - Integration layer and proxy server

Support

For issues, questions, or contributions:

1. Check this README
2. Enable debug mode: cclm config set debug true
3. Review proxy logs
4. Check LM Studio server status

Enjoy using Claude Code with LM Studio! 🚀