Skip to main content

​
Configure

In Cursor Settings → Models:
  • Enable Override OpenAI Base URL and set it to:
  • Paste your sdn_ key into the OpenAI API Key field.
  • Under model names, add a custom model for each catalog id you want to use — e.g. @sdn/gpt-oss-120b, @sdn/glm-5.3, @sdn/deepseek-v4-pro.
  • Click Verify — Cursor sends a probe to /v1/chat/completions and should turn green.

​
Troubleshooting

  • Verify fails? The base URL must end in /v1 and at least one custom model name must be a valid catalog id — Cursor verifies with the model currently selected.
  • Model “not found”? Model names are exact — @sdn/glm-5.3, not glm 5.3 or GLM-5.3-Turbo. Copy ids from the catalog page (older spellings such as sdn-glm-5.3 and glm-5.3 also work).
  • Built-in models stopped working? Expected: overriding the OpenAI base URL points Cursor’s OpenAI slot at Sideren. Toggle the override off to switch back.
Cursor uses its own accounts for some built-in features; the override governs the chat/composer models you add by name. A model you add has to be in your plan (Plans & limits), or run in PAYG mode.

​
Cursor CLI (agent)

Cursor Agent CLI (agent / cursor agent) does not support BYOK or a custom OpenAI-compatible base URL — it only talks to Cursor Cloud under your Cursor account. There is no config file the setup helper can write.
prints the IDE override above and a one-liner for a terminal agent Sideren can configure (hermes, codex, claude-code).