Last checked November 2, 2026. The November 12 date remains proposed, not confirmed. Recheck the linked official sources before changing a production workflow.
Direct answer: OpenAI has proposed November 12, 2026 as the end of its model-supply transition with Cursor, but the final termination date is not yet confirmed and Cursor may end access sooner.[1][2] You do not need to abandon Cursor: for supported local Chat and Agent work, you can add your own OpenAI API key; you can install the separate Codex IDE extension and sign in with ChatGPT; or your organization can connect a compatible AI gateway.[2]
Do this now: Pick one of the three routes below, test it in a non-critical project, and keep a non-OpenAI model enabled as a fallback. Do not describe November 12 as a guaranteed cutoff—the official wording is “proposed.”[1][2]
Quick decision guide
- Choose your own OpenAI API key if you want OpenAI models inside Cursor’s local Chat or Agent and accept separate usage-based API billing.[2]
- Choose the Codex IDE extension if you already have an eligible ChatGPT subscription and want a separate Codex panel inside Cursor.[2][4]
- Choose an AI gateway if your company already manages model access through Azure, Amazon Bedrock, or another compatible gateway.[2]
- Choose another Cursor model if you mainly need Tab, Auto, Cloud/Background Agents, Automations, CLI, API, or SDK workflows, because a personal OpenAI key does not cover those surfaces.[2]
Option 1: Add your own OpenAI API key to Cursor
OpenAI says a personal API key can power supported local Chat and Agent requests in Cursor. API access requires an OpenAI Platform account with billing configured, and these charges are separate from both Cursor and ChatGPT subscriptions.[2]
Setup steps
- Create an API key in the OpenAI Platform dashboard.
- In Cursor, open Cursor Settings → Models.
- Find the OpenAI provider, paste the key, then save or verify it.
- Enable an available OpenAI model.
- Select that model in a local Chat or Agent session and run a small test.[2][3]
Important limits
A custom OpenAI key does not replace OpenAI access across every Cursor feature. OpenAI lists Cursor Tab/autocomplete, Auto routing, Cloud or Background Agents, Automations, Cursor CLI, and Cursor’s API/SDK as unsupported for bring-your-own-key access.[2]
Cursor’s documentation adds two privacy details worth reviewing: custom keys work only with chat models, and requests are routed through Cursor’s servers for final prompt building even though Cursor says the key is encrypted in transit and not persisted after the request.[3] Cursor also says its own Zero Data Retention policy does not apply to your personal provider key, so your provider’s data policy governs that use.[3]
Option 2: Install the Codex IDE extension in Cursor
The Codex IDE extension runs as a separate agent experience inside Cursor. It does not depend on OpenAI models appearing in Cursor’s own model picker, and it can authenticate with an eligible ChatGPT subscription or an OpenAI API key.[2]
Setup steps
- Open the official Codex IDE extension page and choose the Cursor install link.
- In Cursor, open Codex from the Activity Bar. If the icon is hidden, open the Command Palette and run Codex: Open Codex Sidebar.
- Select Sign in with ChatGPT or Use API Key.
- Open a test repository and ask Codex to explain a file before allowing edits.
- Create a Git checkpoint before accepting larger changes.[4]
The extension is independent of Cursor Chat, Agent, Tab, Auto, and Cloud Agents; installing it does not change the model used by those Cursor services.[2]
Option 3: Connect an AI gateway
A gateway is the business-oriented route when model credentials, billing, and access controls are already managed centrally. OpenAI’s guidance names Amazon Bedrock and Azure as examples and also describes custom OpenAI-compatible endpoints.[2]
Setup checklist
- Provision the OpenAI model with your provider.
- Record the endpoint, credentials, and model or deployment name.
- Open Cursor Settings → Models.
- Select the provider integration or enter the custom OpenAI-compatible base URL.
- Add the credentials and enable the model.
- Test a local Chat or Agent request.[2]
Compatibility is not automatic: the endpoint must implement the API format Cursor expects, and some model controls may not pass through a custom gateway.[2] Gateway credentials have the same surface limitation as personal keys—they do not cover Tab, Auto, Cloud/Background Agents, Automations, Cursor CLI, or Cursor’s API/SDK.[2]
Migration checklist before the proposed cutoff
- [ ] List the OpenAI models your team currently selects in Cursor.[2]
- [ ] Note whether each workflow runs in local Chat/Agent or an unsupported surface such as Tab or Cloud Agents.[2]
- [ ] Select API key, Codex extension, gateway, or a non-OpenAI Cursor model for each workflow.[2]
- [ ] Set an API spending limit and alerts if using a personal or organization API account.[2]
- [ ] Test with a small repository that contains no production secrets.
- [ ] Review provider data-handling rules before sending proprietary code.[3]
- [ ] Keep a fallback model enabled.
- [ ] Recheck the official Help Center page because OpenAI says it will publish the confirmed termination date after agreement with Cursor.[2]
What this change does—and does not—mean
OpenAI announced that it intends to wind down the contract that provides OpenAI models to Cursor following Cursor’s change of control, with November 12 as the proposed shutoff date.[1] This is not an announcement that the Cursor editor itself is shutting down.
OpenAI’s Help Center currently says existing OpenAI access can continue during the proposed transition, while warning that Cursor may choose to end access sooner.[2] That uncertainty is why users should migrate before the date rather than waiting for an error message.
FAQ
Is November 12, 2026 the confirmed cutoff?
No. OpenAI calls it a proposed transition or shutoff date and says the confirmed termination date will be shared after it is agreed with Cursor.[1][2]
Can I use my ChatGPT subscription as an API key in Cursor?
No. OpenAI says ChatGPT subscriptions do not include API usage. A personal OpenAI API key is billed through the OpenAI Platform account.[2] The separate Codex IDE extension can instead support sign-in with an eligible ChatGPT subscription.[2][4]
Will my API key keep Cursor Tab working with OpenAI?
No. OpenAI and Cursor both state that custom keys do not power Tab completion; Cursor Tab continues to use Cursor’s built-in models.[2][3]
Does the Codex extension replace Cursor Agent?
No. Codex appears as its own extension and agent panel. It does not change Cursor Chat, Agent, Tab, Auto, or Cloud Agent model selection.[2]
Should a company use a personal key?
Not by default. Teams should first review billing ownership, source-code handling, data-retention requirements, and whether personal keys are allowed by organization policy. Cursor notes that enterprise administrators can restrict personal API keys, so a managed gateway or centrally controlled provider account may be more appropriate.[3]
Is it safe to wait until November 12?
It is safer to test a replacement now because OpenAI says Cursor may end access sooner and the final date remains unconfirmed.[2]
Sources
[1] https://openai.com/index/our-decision-on-cursor-following-its-acquisition-by-spacex — Our decision on Cursor following its acquisition by SpaceX
[2] https://help.openai.com/en/articles/20001506-using-openai-models-in-cursor — Using OpenAI models in Cursor
[3] https://cursor.com/docs/settings/api-keys — Bring your own API key | Cursor Docs
[4] https://developers.openai.com/codex/ide — Codex IDE extension