Ability to add local LLMs to cursor

Feature request for product/service

Cursor IDE

Describe the request

Add Local LLM Providers to Cursor Chat/Agent Model Picker (eg: OpenClaw running Ollama/qwen or other OpenAI-compatible endpoints)

Please add support for local/self-hosted LLMs in Cursor’s chat and agent model selection.

I want to run models on my own GPU (for example via Ollama/OpenAI-compatible servers like OpenClaw + Qwen) and select them directly inside Cursor alongside hosted models.

Requested capabilities (v1):
Add custom model providers/endpoints (local URL, API key/token, model name)
Show local models in the same chat/agent model picker
Support both chat and agent workflows
Per-model settings (context window, temperature, max tokens, tool-calling on/off)
Clear fallback behavior when local model/server is unavailable

Why this matters:
infinite tokens
tighter integration with local computer
Privacy/data residency for sensitive code
Cost control for heavy usage
Flexibility to use best model per task (hosted + local hybrid)

Example use case:
Running qwen via ollama on a local GPU box and choosing that model as the active Cursor chat/agent model from inside Cursor.