Check the settings
- Open Settings → Models and confirm your account allows personal API keys.
- If Override OpenAI Base URL is available, set it to https://vectide.cn/v1 and enter your Vectide key in the OpenAI key field.
- Add an included model ID, such as glm-5.3, rather than a display name.
- Save, open a fresh conversation, select the model and send a short text request.
Check features individually
Cursor documents personal API keys for standard chat models. Features relying on specialized models, such as Tab, are separate. Check agent tools, model validation, endpoint settings and team policies individually.
Cursor routes related requests through its backend. Consider its handling of data and credentials when choosing the client; a personal key does not mean processing stays on your device.
If the connection fails
| Case | Next steps |
|---|---|
| No custom endpoint option | Check version, account policy and Cursor documentation. |
| Model validation fails | Test the same key and model with a minimal HTTP request, then compare client settings. |
| Chat works; tools fail | Check model and client tool support separately. |
| Unexpected usage | Check Vectide usage and Cursor account records separately. |