Cursor’s OpenAI base URL override, for chat with a served model.
Where a request from Cursor goes: the two settings Connect writes, the warden in the middle, then one copy of the model per GPU.
In words:Cursor sends its requests to LM Warden with Override OpenAI Base URL set to the warden and OpenAI API Key holding a warden key. LM Warden checks the key, speaks OpenAI Chat Completions, queues the request by key priority and hands it to the replica router, which keeps each session on the replica that already holds its cache, one model copy per GPU.
Protocol
OpenAI Chat Completions
Support
Local models only
Requests are issued from Cursor’s servers, so this warden’s URL must be reachable from the internet; a LAN or localhost warden will not work. Custom keys only work with chat models: Tab completion keeps using Cursor’s own models, and Agent with Cursor’s models rejects custom keys. Cursor’s current docs no longer describe the Override OpenAI Base URL setting; it is known from Cursor staff forum replies.
Requirements
Cursor’s servers call this URL, so it must be reachable from the internet.
Setup
In the console, Connect writes these files with your warden’s address, your key and the model you picked. Here they are with placeholders.
Chat with a local model
Cursor Settings > Models, then add the model name as a custom model.
Settings
OpenAI API Key: vw_YOUR_KEY
Override OpenAI Base URL: https://your-warden/v1
Custom model: your-served-model-name