> ## Documentation Index
> Fetch the complete documentation index at: https://docs.profclaw.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# LM Studio

> Run local models via LM Studio's built-in OpenAI-compatible server. GUI-based model management with hardware acceleration.

LM Studio provides a user-friendly desktop application for downloading and running local AI models. Its built-in server exposes an OpenAI-compatible API that profClaw connects to.

## Overview

LM Studio is a GUI application for macOS, Windows, and Linux that:

* Downloads models from HuggingFace with one click
* Provides hardware-accelerated inference (Metal, CUDA, CPU)
* Runs a local OpenAI-compatible API server
* Works without an internet connection after model download

## Setup

<Steps>
  <Step title="Download LM Studio">
    Download from [lmstudio.ai](https://lmstudio.ai) and install.
  </Step>

  <Step title="Download a model">
    Open LM Studio, search for a model (e.g., `lmstudio-community/Meta-Llama-3.1-8B-Instruct-GGUF`), and download it.
  </Step>

  <Step title="Start the local server">
    In LM Studio: go to **Local Server** tab, select a model, and click **Start Server**.

    The server runs at `http://localhost:1234` by default.
  </Step>

  <Step title="Configure profClaw">
    ```bash theme={null}
    export LM_STUDIO_BASE_URL=http://localhost:1234/v1
    # Or configure as an OpenAI-compatible endpoint:
    export OPENAI_BASE_URL=http://localhost:1234/v1
    export OPENAI_API_KEY=lm-studio
    ```
  </Step>

  <Step title="Verify">
    ```bash theme={null}
    profclaw doctor --provider openai
    ```
  </Step>
</Steps>

## Environment Variables

<ParamField path="LM_STUDIO_BASE_URL" type="string">
  LM Studio server URL. Defaults to `http://localhost:1234/v1`. Configure via `OPENAI_BASE_URL`.
</ParamField>

## Configuration Example

<Tabs>
  <Tab title=".env">
    ```bash theme={null}
    # LM Studio uses the OpenAI provider with a local base URL
    OPENAI_API_KEY=lm-studio
    OPENAI_BASE_URL=http://localhost:1234/v1
    ```
  </Tab>

  <Tab title="settings.yml">
    ```yaml theme={null}
    providers:
      openai:
        api_key: "lm-studio"
        base_url: "http://localhost:1234/v1"
        default_model: "local-model"
    ```
  </Tab>

  <Tab title="Remote LM Studio">
    ```bash theme={null}
    # Access LM Studio on another machine
    OPENAI_API_KEY=lm-studio
    OPENAI_BASE_URL=http://192.168.1.100:1234/v1
    ```
  </Tab>
</Tabs>

## Usage Examples

<CodeGroup>
  ```bash CLI theme={null}
  # Use whatever model is loaded in LM Studio
  profclaw chat --model local-model "Explain dependency injection"

  # Reference LM Studio model by name
  profclaw chat --model "Meta-Llama-3.1-8B-Instruct-Q4_K_M" "Quick question"
  ```
</CodeGroup>

## Differences from Ollama

| Feature      | LM Studio         | Ollama                     |
| ------------ | ----------------- | -------------------------- |
| Interface    | GUI desktop app   | CLI/headless               |
| Model format | GGUF              | GGUF                       |
| API          | OpenAI-compatible | Custom + OpenAI-compatible |
| Best for     | Desktop users     | Servers, Docker            |
| Auto-start   | No (manual)       | Yes (systemd/launchd)      |

## Notes

* LM Studio's API key field accepts any non-empty string - use `lm-studio` as a placeholder.
* Tool calling support depends on the loaded model - check the model's capabilities.
* LM Studio supports Metal (Apple Silicon), CUDA, and Vulkan acceleration.
* For headless/server deployments, Ollama is recommended over LM Studio.

## Related

* [AI Providers Overview](/ai-providers/overview) - Compare all 37 supported providers
* [Ollama](/ai-providers/ollama) - Recommended local provider for headless deployments
* [Local LLM Guide](/guides/local-llm) - Run profClaw fully offline with local models
* [profclaw provider](/cli/provider) - Add and test providers from the CLI