Cursor

Cursor supports custom API keys for standard chat models and exposes an OpenAI base URL override in its Models settings.

#Configure Cursor manually

1. Create a project key

   Create a key in API keys and keep it available while you configure Cursor. Do not paste the key into project files or Cursor rules.

2. Enter the key in Cursor

   Open Cursor Settings → Models → API Keys. Turn on the OpenAI API key toggle to enable custom API key use, then enter the Fast Inference project key in the OpenAI API key field.

3. Override the OpenAI base URL

   Turn on the Override OpenAI base URL toggle and set the OpenAI base URL override to:

   textCopy

   https://api.inference.net/v1

4. Add your model

   Add a standard chat model ID from the model catalog to Cursor's model list. Some models already appear in Cursor's list — for those, toggle the model on instead of adding it.

5. Send a test prompt

   Start a new standard chat with the custom model. After the response completes, confirm its spend and tokens in Usage.

#CLI availability

The released fast CLI does not currently register fast cursor on. Manual setup is not tracked by the CLI, and fast cursor off cannot restore your prior Cursor settings.

#Revert the manual setup

Return to Cursor Settings → Models → API Keys. Turn off the OpenAI API key toggle or remove the custom key, turn off the Override OpenAI base URL toggle, and reselect the Cursor-managed model you used previously.

#Troubleshooting

• Requests fail with the custom base URL — include /v1 in the base URL override.
• Cursor rejects the selected model — use a model supported by Cursor's standard chat custom-key flow.
• Chats keep using the old endpoint or model — start a new chat after changing either one.
• If the test prompt fails — test the same key and model with the getting-started request to separate gateway and Cursor configuration issues.