Get Houtini LM running in 5 minutes. This guide covers everything you need: Node.js, LM Studio, and Claude Desktop integration. You'll be analysing code locally before your next coffee break.

What You're Building

By the end of this setup, you'll have a powerful local AI development environment that saves you money and gives you unlimited code analysis. Here's what we'll install:

Node.js

JavaScript runtime for Houtini LM

→

Houtini LM

Your local AI analysis toolkit

→

LM Studio

Local AI model hosting

→

Claude Desktop

Your AI development partner

Total Time: 5 Minutes

Most of this is downloading - perfect time to grab a coffee while LM Studio downloads your model.

Prerequisites

Before we start, make sure you have:

Modern Computer

Windows: Windows 10+ with AVX2 support, 4GB+ GPU VRAM recommended
macOS: Apple Silicon (M1/M2/M3/M4), macOS 13.4+
Linux: Ubuntu 20.04+, x64 with AVX2 support
RAM: 16GB+ recommended (8GB minimum for smaller models)

Desktop Commander

Desktop Commander MCP

Essential companion tool for file operations and local development workflows. Handles reading, writing, and managing your project files.

Get Desktop Commander

Administrator Access

To install Node.js and LM Studio on your machine.

LM Studio

LM Studio

The local LLM runtime that powers Houtini's unlimited analysis and generation. Runs powerful models privately on your machine.

Download LM Studio

Step 1: Install Node.js

Node.js powers Houtini LM. If you already have Node.js installed, you can skip to the next step.

Windows Installation

Download and install the Windows installer from the official Node.js website:

1

Download Node.js

Visit nodejs.org and download the "LTS" version (recommended for most users).

Download Node.js
2

Run the Installer

Run the downloaded .msi file and follow the installation wizard. Keep all default options selected.

3

Verify Installation

Open Command Prompt or PowerShell and verify Node.js is installed:

Command Prompt 
node --version

You should see something like v20.11.0 or similar.

macOS Installation

Download and install the macOS installer from the official Node.js website:

1

Download Node.js

Visit nodejs.org and download the "LTS" version for macOS.

Download Node.js
2

Run the Installer

Run the downloaded .pkg file and follow the installation wizard. Keep all default options selected.

3

Verify Installation

Open Terminal and verify Node.js is installed:

Terminal 
node --version

You should see something like v20.11.0 or similar.

Linux Installation

Install Node.js using your distribution's package manager or from the official website:

Ubuntu/Debian 
curl -fsSL https://deb.nodesource.com/setup_lts.x | sudo -E bash -
sudo apt-get install -y nodejs
CentOS/RHEL/Fedora 
curl -fsSL https://rpm.nodesource.com/setup_lts.x | sudo bash -
sudo yum install -y nodejs

Step 2: Install Houtini LM

Now let's install Houtini LM via npm. This gives you access to all the powerful analysis functions.

1

Install Houtini LM Package

Open your terminal or command prompt and run:

NPM Install Command 
npm install -g @houtini/lm
Installing Houtini LM via NPM

NPM installing Houtini LM and its dependencies

2

Verify Installation

Check that Houtini LM installed correctly:

Verification Command 
houtini-lm --version

You should see the current version number displayed.

Permission Issues on Mac/Linux?

If you get permission errors, you might need to use sudo npm install -g @houtini/lm or set up npm to install packages globally without sudo.

Step 3: Install LM Studio

LM Studio hosts the local AI models that power Houtini LM's analysis functions. It's free and runs entirely on your machine.

1

Download LM Studio

Visit the LM Studio website and download the installer for your operating system:

Download LM Studio
LM Studio Interface

LM Studio's clean interface for managing local AI models

2

Install LM Studio

Run the downloaded installer and follow the installation wizard. It's straightforward - just keep the default options.

3

Download a Model (Our Favourite)

Launch LM Studio and download our recommended model:

Qwen2.5-Coder-32B-Instruct (Recommended)

Our absolute favourite for code analysis. Brilliant at understanding complex codebases and providing practical insights.

🧠 32B parameters 💾 ~20GB download 🚀 Excellent for code

In LM Studio, search for "Qwen2.5-Coder-32B" and download the GGUF version. It's large but worth the wait!

4

Load the Model

Once downloaded, click "Load Model" in LM Studio and select your Qwen2.5-Coder model. Make sure the local server is running on port 1234.

Coffee Break Time!

The Qwen2.5-Coder model is ~20GB, so this is perfect time for a coffee break. The download runs in the background whilst you can continue with other work.

Step 4: Configure Claude Desktop

The final step is adding Houtini LM to your Claude Desktop configuration so you can use it directly in your conversations.

Windows Configuration

Add Houtini LM to your Claude Desktop configuration file:

1

Open Configuration File

Navigate to %APPDATA%\Claude\claude_desktop_config.json and open it in your text editor.

2

Add Houtini LM Configuration

Add this JSON configuration to your claude_desktop_config.json file:

claude_desktop_config.json 
{
  "mcpServers": {
    "houtini-lm": {
      "command": "houtini-lm",
      "args": [],
      "env": {
        "LLM_MCP_ALLOWED_DIRS": "C:\\dev,C:\\projects,C:\\work"
      }
    }
  }
}

macOS Configuration

Add Houtini LM to your Claude Desktop configuration file:

1

Open Configuration File

Navigate to ~/Library/Application Support/Claude/claude_desktop_config.json and open it in your text editor.

2

Add Houtini LM Configuration

Add this JSON configuration to your claude_desktop_config.json file:

claude_desktop_config.json 
{
  "mcpServers": {
    "houtini-lm": {
      "command": "houtini-lm",
      "args": [],
      "env": {
        "LLM_MCP_ALLOWED_DIRS": "/Users/yourusername/dev,/Users/yourusername/projects"
      }
    }
  }
}

Replace "yourusername" with your actual macOS username.

Linux Configuration

Add Houtini LM to your Claude Desktop configuration file:

1

Open Configuration File

Navigate to ~/.config/Claude/claude_desktop_config.json and open it in your text editor.

2

Add Houtini LM Configuration

Add this JSON configuration to your claude_desktop_config.json file:

claude_desktop_config.json 
{
  "mcpServers": {
    "houtini-lm": {
      "command": "houtini-lm",
      "args": [],
      "env": {
        "LLM_MCP_ALLOWED_DIRS": "/home/yourusername/dev,/home/yourusername/projects"
      }
    }
  }
}

Replace "yourusername" with your actual Linux username.

Important: Set Your Allowed Directories

Change the LLM_MCP_ALLOWED_DIRS paths to match your actual project directories. Houtini LM can only access files within these specified directories for security.

Step 5: Restart Claude Desktop

The final step - restart Claude Desktop to load your new configuration:

1

Close Claude Desktop

Completely quit Claude Desktop (not just minimize - actually quit the application).

2

Restart Claude Desktop

Launch Claude Desktop again. It will load your new Houtini LM configuration.

Verify Everything Works

Let's make sure your setup is working perfectly. Try these quick tests in Claude Desktop:

Test 1: Health Check

In Claude Desktop, type: houtini-lm:health_check

Expected: You should see connection status and model information.

Test 2: List Functions

Try: houtini-lm:list_functions

Expected: You should see all available Houtini LM functions.

Test 3: Analyse a File

Try: Use houtini-lm to analyse this file: [path to any code file]

Expected: You should get detailed code analysis and insights.

🎉 You're All Set!

If all tests pass, congratulations! You now have unlimited local AI code analysis. Start with analysing a single file to see what Houtini LM can do.

Troubleshooting

Hit a snag? Here are solutions to the most common setup issues:

Houtini LM functions not showing up in Claude Desktop

Claude Desktop isn't recognising the Houtini LM MCP server.

  • Check your JSON syntax in claude_desktop_config.json - even a missing comma breaks everything
  • Ensure you've completely restarted Claude Desktop (quit and relaunch)
  • Verify the houtini-lm command works in terminal: houtini-lm --version
  • Check that Node.js is in your PATH by running node --version
LM Studio connection errors

Houtini LM can't connect to LM Studio for local model hosting.

  • Ensure LM Studio is running and has a model loaded
  • Check that LM Studio's local server is running on port 1234
  • Try restarting LM Studio and reloading your model
  • In LM Studio settings, ensure "Local Server" is enabled
Permission denied errors on file analysis

Getting permission errors when trying to analyse files.

  • Check that your file path is within the directories listed in LLM_MCP_ALLOWED_DIRS
  • Use absolute paths: C:\project\file.js not .\file.js
  • Ensure the file exists and you have read permissions
  • Update LLM_MCP_ALLOWED_DIRS to include your project directories
npm install fails or permission errors

Can't install Houtini LM via npm due to permission issues.

  • On Mac/Linux: Try sudo npm install -g @houtini/lm
  • Better solution: Configure npm to install packages globally without sudo
  • Check npm configuration: npm config get prefix
  • Ensure you're using a recent version of Node.js (18+ recommended)
Model download is very slow or fails

The Qwen2.5-Coder model download is taking forever or failing.

  • The model is ~20GB - downloads can take 30+ minutes on slower connections
  • Try a smaller model first like "Qwen2.5-Coder-7B" for testing
  • Ensure you have at least 25GB free disk space
  • Try downloading during off-peak hours if your internet is slow

Next Steps

You're all set up! Here's what to explore next:

Try Your First Analysis

Start with analyzing a single file - it's the perfect introduction to Houtini LM's capabilities

Explore Project Structure

Get a beautiful overview of your entire codebase with count_files