OpenAI plans to remove its models from Cursor after SpaceX deal
A proposed November 12 cutoff shows why model access, credentials and evaluations cannot remain bundled inside one coding editor.
The cutoff
OpenAI said on August 28 that it intends to end the contract supplying its models through Cursor. The proposed cutoff is November 12, 2026, and OpenAI says future models will not be provided to Cursor.
The portability gap
Bringing an OpenAI API key can preserve local Chat and Agent access. It does not cover Cursor Tab, Auto, cloud or background agents, automations, CLI, API or SDK, so migration is not a simple credential change.
What we would do
We would keep Cursor where its workflow earns the cost, but move keys, routing, usage records and evaluations under company control. For important tasks, we would maintain a second agent harness as a tested fallback.
OpenAI’s Cursor cutoff is not primarily a model story; it is a dependency-management failure exposed by a change of ownership. A team can keep the same repository, editor and prompts yet still inherit a mandatory migration project because the commercial relationship above its tooling changed. We would not abandon Cursor tomorrow, but we would stop treating its bundled model picker as infrastructure. We would inventory workflows that depend on OpenAI behaviour, rerun representative tasks through a company-controlled API key or AI gateway, and preserve a second harness for the jobs that matter. We would also pause new critical automations on Cursor features that cannot use bring-your-own credentials, including Tab, Auto, cloud and background agents, automations, CLI, API and SDK paths. The proposed November 12 deadline is enough time to test, not enough time to postpone. The operator lesson is blunt: choose coding tools for their workflow, but own the model access, evaluation set, budgets and fallback path yourself.