Cursor

Cursor reaches Astorias through its OpenAI settings. You paste an ast_ key as the OpenAI key, point the OpenAI base URL at Astorias, and add catalog ids as custom models.

Setup

  1. 1

    Create an API key

    Organization admins create keys at APIs. Copy the ast_ key when it is shown.
  2. 2

    Open the key settings

    Open Cursor Settings, go to Models, and expand API Keys.
  3. 3

    Paste the key

    Turn on OpenAI API Key and paste the ast_ key.
  4. 4

    Override the base URL

    Turn on Override OpenAI Base URL and set it to the URL below. Leave off /chat/completions. Cursor adds the path.
    https://api.astorias.ai/v1
  5. 5

    Add a model

    Click Add model and paste an id from the list below, once per model you want. Then pick it in the chat model picker.
    • qwen/qwen3.8-27b
      Qwen 3.8 27B256K contextImageToolsReasoning
    • z-ai/glm-5.3-flash
      GLM 5.3 Flash1M contextImageToolsReasoning

What to know

Chat and Agent only

The key applies to local Chat and Agent. Tab completion, Auto, Cloud Agents, and the Cursor CLI keep using Cursor's own models.

The override covers every OpenAI model

While the override is on, Cursor sends OpenAI-family requests to Astorias, including models picked from Cursor's built-in list. Turn the override off to use those again.

The key passes through Cursor

Cursor forwards the key through its servers on each request to build the prompt. Treat it like any production key and revoke it if a machine is lost.

Credits and limits

Cursor requests spend organization credits and share the key's rate limit. A 401 means the key is wrong or revoked. A 400 on a tool call usually means the model does not list tools. See Errors.