Cursor

Cursor works with codexkey in OpenAI-compatible mode, using your own key.

Connecting

1. Open Cursor → Settings → Models.
2. Turn on OpenAI API Key and paste your codexkey key.
3. Turn on Override OpenAI Base URL and enter https://api.codexkey.online/v1.
4. Click Add model and type the Cursor model name from the table below.
5. Click Verify and pick the model in the chat.

Model name for Cursor

With a third-party key Cursor does not accept its own model names, so Claude and GPT have a separate name for Cursor: for example, cursor5 for Claude Opus 5. Models from other developers (Gemini, DeepSeek, Qwen and others) are added under the name from the API.

The table sometimes lags behind the catalog. If a Claude or GPT model you need is missing, support can tell you its name.

Cursor limitations

• A custom Base URL applies to all models: built-in ones (Composer, Auto) may not work while it is on.
• You cannot switch models in the middle of a chat: add the ones you need in advance.
• If Verify fails or the Base URL switch is missing, install the Cline or Continue extension in Cursor (see VS Code and others).

When something goes wrong

Model not found

Type the name from the table exactly, with no spaces or stray characters. If a model has no Cursor name, take its ID from GET /v1/models.

Authentication errors

The key starts with sk_cdx_, has no spaces or line breaks, and stays active in the API Keys section.