Connect LM Studio via Custom
Prepare LM Studio
Section titled “Prepare LM Studio”- Install LM Studio and consult the local server and OpenAI-compatible API.
- Download and load a model with tool support. Confirm the identifier provided by the server, not just the name of the downloaded file.
- Enable the local server and check the port shown by the application. The usual setup uses
localhost:1234, but your installation may differ. - Keep the service limited to the device; do not expose the server publicly.
Configure Browser Mate
Section titled “Configure Browser Mate”- Open Settings → LLM Model and select Custom (OpenAI-compatible).
- Enter the exact identifier of the served model.
- Fill in the full URL, for example
http://localhost:1234/v1/chat/completions. Browser Mate does not append the path in Custom. - Fill in API key: the current runtime requires a non-empty value for Custom, even when the local server does not require authentication.
- If you enabled authentication on the server, use the real token defined in LM Studio. If the server does not require authentication and accepts an arbitrary Bearer header, you can use a local placeholder such as
lm-studio-local. That placeholder does not protect the server. - Save and test the connection. Start with text, then check tools and images separately.
Compatibility and errors
Section titled “Compatibility and errors”This route uses Chat Completions, tools and streaming compatible with OpenAI. A text-chat-only or Responses API is not equivalent. “Unknown capability” indicates missing information, not a successful test.
On connection errors, check the running server/model, URL, port, token and allowed origins. Do not solve CORS by exposing the service to all origins. The guide is based on the runtime and the documented interface; no call was made to your local server.
