OpenAI is pulling its models from Cursor on November 12
After SpaceX's buyout, current GPT models keep working in Cursor until the cutoff — but bring-your-own-key won't power Tab, autocomplete, or Cloud Agents.
Copy markdown
The cutoff: November 12
OpenAI proposed Nov 12, 2026 as the day it stops serving its models to Cursor, after SpaceX's ~$60B acquisition of the editor. Models already wired in keep working until then; future GPT releases won't be offered under the current deal. You have roughly two months to plan a migration.
BYOK is a patch, not a fix
You can plug a personal OpenAI API key into Cursor for Chat and Agent tasks — but it won't drive Tab, autocomplete, Auto/Cloud/Background Agents, Automations, or the CLI and SDK. The surfaces you lean on daily are exactly the ones a personal key can't reach.
You'll pay OpenAI separately
ChatGPT Plus and Pro don't include API credits, so any bring-your-own-key usage bills per token on top of your Cursor plan. Budget for a second invoice if you want to keep GPT in the loop.
The clean escape: Codex extension
OpenAI's Codex IDE extension installs directly inside Cursor (it's a VS Code fork), giving you OpenAI's coding agents independent of the model agreement. It's the one path that keeps full agent behavior running on GPT.
Why it happened
OpenAI says it 'cannot be confident that SpaceX will use our technology within our terms of service.' Cursor's co-founder counters that OpenAI is only ~5% of the editor's model traffic — and Anthropic kept Claude available in Cursor.
The default move: switch models
Claude and other providers stay live in Cursor, and at 5% of traffic most workflows won't feel GPT's exit. If GPT-6 Astra isn't central to how you build, doing nothing is a legitimate option.