Connector for LM Studio Local Inference Engines

This MCP service acts as a middleware layer, linking Claude's capabilities to arbitrary local Large Language Models (LLMs) managed within your LM Studio setup.

Core Functionality

This bridge grants Claude the ability to interface with your local LLM deployments running via LM Studio, encompassing:

• An enumeration function to retrieve the roster of all models currently accessible within LM Studio.
• The capability to prompt your local LLMs for direct text synthesis.
• Full support for turn-based chat interactions utilizing your selected local models.
• A status verification utility to instantly confirm operational connectivity with the LM Studio API server.

Prerequisites for Operation

To utilize this tool, the following components must be in place:

• Access to the Claude Desktop application, configured for MCP extensions.
• A functional installation of LM Studio running on the local machine, with its integrated API server activated.
• A runtime environment compatible with Python version 3.8 or newer.

Rapid Deployment Guide (Recommended)

For Unix-like Systems (macOS/Linux):

1. Obtain the source code repository:

git clone https://github.com/infinitimeless/claude-lmstudio-bridge.git
cd claude-lmstudio-bridge

1. Execute the setup utility:

chmod +x setup.sh
./setup.sh

1. Adhere to the on-screen directions provided by the setup script for Claude Desktop integration.

For Windows Systems:

1. Clone the repository locally:

git clone https://github.com/infinitimeless/claude-lmstudio-bridge.git
cd claude-lmstudio-bridge

1. Launch the batch setup file:

setup.bat

1. Follow the configuration steps presented by the script to finalize Claude Desktop linkage.

Manual Configuration Procedure

If automated setup is bypassed:

1. Establish a dedicated Python virtual environment (recommended practice):

python -m venv venv
source venv/bin/activate  # For Windows: venv\Scripts\activate

1. Install necessary dependencies from the manifest file:

pip install -r requirements.txt

1. Configure the external server linkage within Claude Desktop:
2. Navigate to Claude Desktop's settings panel.
3. Locate the 'MCP Servers' configuration zone.
4. Define a new server entry with these parameters:
   • Identifier Name: lmstudio-bridge
   • Execution Program: /bin/bash (Linux/macOS) or cmd.exe (Windows)
   • Execution Arguments:
   • Unix: /full/absolute/path/to/claude-lmstudio-bridge/run_server.sh
   • Windows: /c C:\full\absolute\path\to\claude-lmstudio-bridge\run_server.bat

Interaction Methods in Claude

Once the bridge is successfully initialized, Claude can issue the following types of requests:

1. Verification of the LM Studio link status:

Can you confirm my LM Studio API service is reachable?

1. Requesting the library of loaded models:

Display a list of all models currently loaded in my local LM Studio environment

1. Invoking text synthesis via a local model:

Compose a brief, evocative stanza concerning the arrival of spring, utilizing my local inference engine.

1. Submitting a dialogue turn for processing:

Query my locally hosted LLM: "Enumerate the principal architectural concepts underpinning transformer networks in ML."

Troubleshooting Guidance

Resolving Connection Failures to LM Studio

Use the included diagnostic script for initial connection assessment:

python debug_lmstudio.py

For more comprehensive diagnostic output:

python debug_lmstudio.py --test-chat --verbose

Common Problems Encountered

Error: "Inability to establish contact with the LM Studio API endpoint" - Ensure LM Studio is actively running. - Double-check that the API server feature is toggled ON within LM Studio's preferences (Settings > API Server). - Validate that the configured port (default is 1234) aligns with the setting in your environment configuration file (.env).

Error: "No language models are currently instantiated" - Open LM Studio and load the desired model into memory. - Confirm that the model has successfully initialized its operational state.

Error: "MCP package dependency missing" - Attempt a clean reinstallation of required packages: pip install "mcp[cli]" httpx python-dotenv - Verify that the Python interpreter version meets the minimum requirement (3.8+).

Error: "Claude cannot locate the bridge service" - Scrutinize the server registration details within Claude Desktop. - Confirm the file path specified for run_server.sh or run_server.bat is absolute and correct. - On Unix systems, ensure the execution bit is set: chmod +x run_server.sh

Configuration Customization

Behavioral aspects of the connector can be tuned by establishing a .env file in the root directory containing these variables:

LMSTUDIO_HOST=127.0.0.1
LMSTUDIO_PORT=1234
DEBUG=false

Setting DEBUG=true activates extensive, verbose logging, aiding in deep diagnostic work.

Software License

Issued under the MIT License.