LM Studio Provider for VS Code

Use local LM Studio models in VS Code chat through the LanguageModelChatProvider API.

This extension discovers chat-capable models from LM Studio, exposes them to VS Code under the lmstudio vendor, and forwards chat requests to LM Studio's OpenAI-compatible /v1/chat/completions endpoint.

Features

• Automatic model discovery from LM Studio
• Uses LM Studio context metadata instead of hardcoded placeholder limits
• Filters out embedding-only models from chat registration
• Refreshes available models on demand with a cached fallback
• Includes built-in diagnostics for provider and model visibility

Requirements

• VS Code 1.90.0 or later
• LM Studio running locally or on a reachable host
• LM Studio API enabled, defaulting to http://localhost:1234
• GitHub Copilot Chat installed in VS Code

Installation

From source

npm install
npm run compile

Then press F5 in VS Code to launch an Extension Development Host.

As an installed extension

Package and install the extension:

npx vsce package

Then install the generated .vsix in VS Code.

Configuration

Set the LM Studio base URL in your VS Code settings if you are not using the default host:

{
  "lmstudio.apiBase": "http://localhost:1234",
  "lmstudio.enableExperimentalAgentMode": false,
  "lmstudio.requestTimeoutSeconds": 120
}

http://localhost:1234 is the built-in default. If you use a different host, IP, or port such as http://localhost:12345, set lmstudio.apiBase directly or use LM Studio: Connection Settings and the extension will use that saved value.

lmstudio.enableExperimentalAgentMode is off by default. This keeps LM Studio models available for normal chat, but hides them from agent/autopilot mode unless you explicitly opt in. Local runtimes like LM Studio and Ollama can loop, unload, or stall during tool-heavy agent turns, so the extension now treats agent mode as experimental.

Usage

1. Start LM Studio.
2. Make sure the LM Studio local server is enabled.
3. Open VS Code with this extension installed.
4. Open chat and pick an LM Studio model from the model selector.

For normal chat, nothing else is required. If you want LM Studio models to appear in agent/autopilot mode too, enable lmstudio.enableExperimentalAgentMode and fully restart VS Code.

If you changed the extension manifest or installed a new .vsix, fully restart VS Code instead of using Reload Window.

Commands

• LM Studio: Run Diagnostics
• LM Studio: Test Available Models
• LM Studio: Connection Settings

The Manage LM Studio... button in VS Code model management opens the same connection settings panel.

Troubleshooting

Check the LM Studio API

curl http://localhost:1234/api/v1/models

You should see LM Studio's model inventory. Chat models should have type: "llm". Embedding models are intentionally ignored by this extension.

Run diagnostics inside VS Code

Use LM Studio: Run Diagnostics from the Command Palette. The output shows:

• whether the extension is active
• which models the extension discovered from LM Studio
• which models VS Code currently sees under the lmstudio vendor

Models do not appear in the picker

• Confirm LM Studio is reachable at the configured URL.
• Confirm you restarted VS Code after installing or updating the extension.
• Confirm the diagnostics command lists the expected models.
• If diagnostics finds models but the picker does not update, close all VS Code windows and launch VS Code again.

Agent mode loops or gets stuck

• Leave lmstudio.enableExperimentalAgentMode off unless you specifically want to test local tool calling.
• If you do enable it, prefer models that report tool-use capability in LM Studio and keep lmstudio.requestTimeoutSeconds high enough for your hardware.
• If LM Studio or Ollama repeatedly loops in agent/autopilot mode, switch back to normal chat for local models. This is usually a local runtime/tool-calling limitation, not a connection problem.

Development

npm install
npm run compile
npm run watch

Project structure

• src/extension.ts registers the provider and commands
• src/lmstudio-provider.ts discovers LM Studio models and serves chat responses
• package.json defines the extension manifest and contributions

License

MIT