Using Cursor with Local LLM (Osaurus)¶

Use Cursor with a local LLM served by Osaurus so you can code without internet (e.g. on a plane). Osaurus exposes an OpenAI-compatible API on http://127.0.0.1:1337.

Prerequisites¶

1. Osaurus installed and running (e.g. osaurus serve --port 1337 or start from the app).
2. At least one model loaded in Osaurus (e.g. a small code model).

Cursor configuration¶

Option A: Cursor Settings UI (recommended)¶

1. Open Cursor Settings (gear icon in the bottom-left).
2. Go to Models.
3. Under OpenAI API Keys (or Add model / custom provider):
4. Override Base URL: http://127.0.0.1:1337/v1
5. API Key: any placeholder (e.g. not-needed or local) — Osaurus often does not require a key.
6. Choose the model that matches what Osaurus is serving (e.g. the model name shown in Osaurus).
7. Restart Cursor if the model list or base URL does not update.

Option B: Workspace settings¶

This repo’s .vscode/settings.json includes:

"cursor.openai.baseUrl": "http://127.0.0.1:1337/v1"

If your Cursor version respects this key, the override is applied for this workspace. Otherwise use Option A.

Before going offline¶

1. Start Osaurus and load your model.
2. In Cursor, set the base URL and model as above.
3. Send a test message in Cursor Chat to confirm the local model responds.

Notes¶

• Same base URL (http://127.0.0.1:1337/v1) is used by this project’s AI migration tools via SL_LLM_BASE_URL (see LLM-based translation).
• Some Cursor features (e.g. certain tab completions) may behave differently with a local model.