Connect Cursor
Use Cursor's custom OpenAI-compatible API path with Primex. Start with the model list in Cursor Setup.
Configure
Open Cursor's model/API-key settings and configure its custom OpenAI base URL. Control names can vary with the installed version.
| Setting | Value |
|---|---|
| OpenAI base URL override | https://gateway.primexos.com/v1 |
| API key | Your Primex key, entered locally |
| Custom model | The exact compatible offering ID selected in Setup |
Enable the custom key and model where required by your version. The OpenAI API format describes the request path; compatible models can have other creators.
Verify chat and the features you use
- Open a disposable project and select the configured model.
- Ask for a short text reply and confirm that it completes.
- Find the request in Usage.
- If you depend on agent tools, ask it to read a harmless local file and complete the tool-result turn. A plain chat request does not verify that feature.
Cursor features such as autocomplete, indexing or specialized agent behavior may use their own models or require a Cursor plan. A custom API key is not a guarantee that every Cursor feature routes through Primex. Review the current Cursor API-key documentation.
Common fixes
- Wrong path: use one
/v1in the configured OpenAI base URL. - Model not found/unavailable: copy the exact ID from Setup and refresh the list.
- Chat works, agent fails: check tools, schemas and protocol support for the selected model and the installed client's actual request.
- Invalid parameters: remove optional sampling settings the model does not support.
- Authentication, balance or limits: follow Errors & Limits.
To switch away, remove the Primex override/custom key and choose the intended provider/model in Cursor. Revoke an exposed key from Primex API Keys.
Need help connecting? Report a setup issue.