LM Studio MCP Bridge
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| LM_HOST | No | The host of the LM Studio server | localhost |
| LM_PORT | No | The port of the LM Studio server | 1234 |
| LM_API_TOKEN | No | Optional API token for authentication |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| query_local_llmA | Standard: Use this for STARTING a new conversation or one-off tasks. Supports Vision, JSON Schema, and Reasoning. Do not use this for follow-ups if you have a response ID. |
| query_local_llm_statefulA | Follow-up: Use this ONLY for continuing an existing conversation when you have a |
| analyze_local_imageC | Vision: Privacy-focused image analysis using local vision models (Llava, Moondream, etc). |
| query_local_fileC | Privacy-First: Reads a local file and queries the local LLM about its contents. |
| search_local_docsB | Local RAG: Performs semantic search across a directory using local embeddings. |
| get_local_embeddingsC | Generate vector representations of text using a local model. |
| get_system_healthB | Check bridge machine and LM Studio host resource usage. |
| list_local_modelsC | List loaded and available models. |
| load_local_modelB | Load a specific model into LM Studio memory. |
| unload_local_modelB | Unload a model instance to free up memory/VRAM. |
| check_server_statusB | Verify LM Studio API connection and health. |
| get_lm_link_statusA | Mesh & Network: Display the status of LM Link and discovered devices on the mesh. |
| manage_lm_linkA | Mesh & Network: Enable/Disable LM Link or set the local device name. |
| set_preferred_lm_link_deviceC | Mesh & Network: Set the preferred LM Link device for remote model resolution. |
| list_files_in_directoryB | List files in a specific local directory. |
| read_file_contentB | Directly read the content of a local file. |
| analyze_local_image_asyncB | Async Vision: Start a background image analysis task. Returns a Task ID. |
| get_bridge_task_statusB | Check the status and result of a background vision task. |
| get_bridge_configB | Retrieve current bridge environment and network configuration. |
| lms_statusB | CLI: Show the overall health of the LM Studio daemon and server. |
| lms_lsB | CLI: List models currently available on disk. |
| lms_psB | CLI: List models currently loaded in memory. |
| lms_getC | CLI: Search for or download models from LM Studio Hub. |
| lms_importC | CLI: Import a model file into LM Studio. |
| lms_server_controlD | CLI: Manage the local inference server. |
| lms_load_cliC | CLI: Advanced model loading with granular control. |
| lms_log_snapshotB | CLI: Capture a snapshot of current logs for a brief duration. |
| lms_runtime_controlC | CLI: Manage and update the inference runtime engines. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 28 tools
Multiple tools have overlapping purposes: get_system_health, check_server_status, and lms_status all report health status; list_local_models, lms_ls, and lms_ps all list models in different forms. The distinction between query_local_llm and query_local_llm_stateful is clear but other boundaries (e.g., lms_server_control vs lms_runtime_control) are fuzzy.
Naming conventions are mixed: most tools use snake_case (get_system_health, query_local_llm), but a subset uses an lms_ prefix (lms_status, lms_ps, lms_get). Verb choices vary inconsistently across similar actions (list vs get vs check vs read), and the async variant `analyze_local_image_async` does not follow a systematic pattern.
With 28 tools, the set is bloated. Many tools duplicate functionality or could be consolidated (e.g., multiple health check tools, multiple model listing tools). A more streamlined set of around 15-18 core tools would better align with the bridge's purpose.
The tool surface covers core workflows: model lifecycle, inference, vision, RAG, embeddings, file access, and network mesh. However, there are notable gaps such as no explicit tool to delete/unload models (only load/unload by model instance) and no way to cancel or manage background tasks beyond checking status. The redundancy in health/status commands also creates confusion but not missing functionality.