OpenAI to end Cursor partnership after SpaceX acquisition: What changes for developers?

Lisa Ernst · 05.09.2026 · AI & Development · 10 min.

OpenAI intends to end its direct model partnership with Cursor following the AI code editor's acquisition by SpaceX. On August 28, 2026, OpenAI announced it would let the contract for providing its models to Cursor expire. The planned shutdown date mentioned was <strong>November 12, 2026</strong>.

It is important to distinguish: The frequently searched phrase <em>openai drops spacex cursor partnership</em> sounds like a complete break. In reality, the decision initially affects the contract through which Cursor directly integrates OpenAI models into its own functionalities. Developers can continue to use OpenAI models through their own API access, the Codex IDE extension, or compatible gateways.

As of September 5, 2026, November 12 is also a <strong>proposed</strong> deadline and has not yet been confirmed as the final contract end date. Therefore, now is the right time for teams to check critical Cursor workflows for dependencies on OpenAI models.

Source: Simple Icons / SpaceX

In Brief

• OpenAI plans to withdraw: The direct contract for providing OpenAI models in Cursor is to be terminated.
• November 12, 2026 is not final yet: OpenAI calls the date proposed and will announce the final date after consultation with Cursor.
• Cursor is not disappearing: The editor and its remaining models, as well as Cursor's own features, will fundamentally remain.
• OpenAI remains technically usable: For local chat and agent functions, your own OpenAI API key or a compatible gateway can be used.
• Codex is a separate route: The Codex IDE extension runs within Cursor independently of the Cursor model selector.
• Not every Cursor feature supports BYOK: Tab autocompletion, auto-routing, cloud and background agents, automations, CLI, and API/SDK cannot use their own OpenAI key.

What happened between OpenAI, Cursor, and SpaceX?

The current situation is the result of several steps within a few months. Cursor had already established a technical relationship with SpaceX before the acquisition. In April 2026, the company announced a model training partnership with SpaceXAI, stating it would use Colossus infrastructure for this. In August, Cursor subsequently confirmed it had been officially acquired by SpaceX.

Why is OpenAI withdrawing the direct partnership?

OpenAI explicitly attributes the decision to the change of ownership. The company states that its individual contract with Cursor contains a limited window for termination after a change of control. OpenAI also explains that, based on its experiences with Elon Musk's companies, it cannot be sufficiently sure that its technology will be used within the agreed terms of use.

This explanation is <strong>OpenAI's position</strong>. It should not be interpreted as an independently determined breach of contract by SpaceX in this specific Cursor relationship. For developers, the operational consequence is most relevant: OpenAI states it wants to make the termination effective as late as possible, but at the same time <strong>no longer provide future models via the Cursor contract</strong>.

Source: simpleicons.org

OpenAI is not fundamentally ending all use of its models in Cursor. The direct contractual model provision to Cursor is affected; own API accesses and Codex remain separate paths.

What does the SpaceX acquisition mean for Cursor?

Cursor itself presents the acquisition as an acceleration of its model strategy. The company had already stated in April that it intended to expand its training activities with the help of SpaceXAI infrastructure. With the change of ownership, Cursor is thus moving more strongly into an ecosystem that controls its own models and its own computing infrastructure.

However, this does not mean that Cursor will exclusively use SpaceX models from now on. Cursor's documentation continues to list models and integrations from various providers. Which models will be offered long-term and how they will be routed into features like Agent, Auto, or Cloud Agents may change. Therefore, for teams, the question <strong>which function is tied to which model provider</strong> is more important than just whether a specific model name appears in the selection menu.

Source: simpleicons.org

Cursor has been part of SpaceX since August 2026. The editor will remain, but its model supply will increasingly depend on alternative providers and its own models following the announced OpenAI withdrawal.

What specifically changes for developers?

The most important point is: <strong>OpenAI usage in Cursor will not automatically become impossible.</strong> However, the access method, billing, and in some functions, the technical scope will change. OpenAI names three alternatives for developers who wish to continue using its models within Cursor.

Option 1: Use your own OpenAI API key in Cursor

For many individual developers, BYOK, i.e., <em>Bring Your Own Key</em>, is the most direct replacement. According to OpenAI and Cursor, the key is entered under <strong>Cursor Settings &gt; Models</strong>. Then, supported OpenAI models can be selected for local chat and agent sessions.

This leads to two common misunderstandings. First, a ChatGPT subscription does not automatically include API usage. API requests are billed separately through the OpenAI API account. Second, your own key does not replace the entire Cursor infrastructure: Cursor Tab and Autocomplete, Auto-routing, Cloud and Background Agents, Automations, Cursor CLI, and Cursor API and SDK continue to use models that Cursor provides or routes itself.

Data privacy with BYOK: Don't just assume the same rules apply

Cursor explicitly points out that its Zero-Data-Retention policy does not automatically apply when using your own API keys. Data processing then depends on the selected provider. Cursor also explains that the API key is routed through Cursor's servers for final prompt creation, transmitted encrypted, and not stored permanently.

For companies, this is an important architectural point: switching from integrated model provision to BYOK changes not only the bill but possibly also data privacy, logging, and compliance assumptions. Enterprise administrators can also block personal API keys in team settings.

Option 2: Use Codex directly as an IDE extension in Cursor

OpenAI explicitly names the Codex IDE extension as a second path. It runs within Cursor but is technically separate from Cursor's own model selector. Developers log in with either a suitable ChatGPT subscription or an OpenAI API account and then work through their own Codex panel.

This is particularly interesting for teams that want to use OpenAI for agentic coding without tying their entire Cursor workflow to direct OpenAI integration. However, the extension does not replace Cursor Chat, Agent, Tab, or Auto. Those who use these features should therefore test them separately. You can also find a more detailed analysis of Codex in our Zerlo guide to OpenAI Codex.

Option 3: Connect OpenAI via Azure, Amazon Bedrock, or a Gateway

For companies with central cloud and cost management, a gateway might make more sense than individual API keys. OpenAI lists Amazon Bedrock, Azure, and OpenAI-compatible gateways among others. Depending on the provider, Cursor can be connected via its own settings or a compatible base URL.

Source: simpleicons.org

Amazon Bedrock is mentioned by OpenAI as a possible gateway path. For teams, this can leverage existing IAM, billing, and governance structures, provided the desired OpenAI model is available and compatible with Cursor.

The advantage lies in central governance: credentials, budgets, and model approvals can be controlled via existing cloud processes. The disadvantage is additional integration complexity. A gateway must support the API format that Cursor expects, and not every model-specific setting will necessarily be passed through.

Source: simpleicons.org

Azure can also serve as a managed access path. The crucial aspect is that the desired model is provisioned within your own Azure setup and Cursor supports the respective provider configuration.

As with your own API key, the same applies to gateways: credentials only apply to the local chat and agent paths supported by Cursor. Anyone using Cloud Agents, Automations, Tab, or Auto cannot simply switch these functions to the same gateway access.

Should developers switch from Cursor now?

For most users, there is no objective reason to leave Cursor immediately solely because of the OpenAI announcement. The editor will continue to work, and several alternative model paths exist. A switch becomes more sensible if a team is heavily dependent on a combination of a <strong>specific OpenAI model and Cursor's proprietary features</strong> that cannot be replicated with BYOK, Codex, or a gateway after the contract ends.

This is precisely why the decision should be based on a workflow test and not on the provider's name. A coding agent can select different files, call tools differently, produce longer or shorter changes, and interpret tests differently with another model on the same repository. Such deviations are more important for productive teams than a general benchmark.

What teams should test before November 12th

1. Inventory OpenAI dependencies: Note which Cursor features and which specific models are used in daily workflows.
2. Test BYOK separately: Test local chat and agent scenarios with your own API key and record actual API costs.
3. Test Codex as an independent path: Compare tasks such as refactoring, testing, bug fixing, and repository-wide changes.
4. Identify Cursor-specific features: Mark anything that requires Tab, Auto, Cloud Agents, Background Agents, Automations, CLI, or API/SDK.
5. Compare alternative models: Use a fixed set of representative tasks instead of subjective single prompts.
6. Re-evaluate data privacy and compliance: BYOK and gateways may have different data processing rules than the previous integrated supply.
7. Check enterprise policies: Verify if personal API keys are allowed at all within the organization.
8. Document fallback: Determine which model or provider path will be used if Cursor terminates OpenAI access before the proposed date.

The bigger lesson: Think of AI code editors and model providers separately

The conflict highlights a structural risk in modern AI development environments. An editor can remain stable while model contracts, routing, prices, or availability change underneath. For development teams, it is therefore worthwhile to treat three levels separately: <strong>editor</strong>, <strong>model access</strong>, and <strong>agentic runtime</strong>.

Those who keep prompts, tests, repository rules, and acceptance criteria model-neutral can more easily handle a provider change. Especially for security-critical or business-critical codebases, an alternative model should not be tested only on the day of an outage. The announced transition period until November offers a concrete time window for this.

FAQ

Conclusion

Following the SpaceX acquisition, OpenAI is withdrawing from the <strong>direct model partnership with Cursor</strong>, but not entirely from the Cursor ecosystem. The proposed November 12th, 2026, is a transition milestone, not an immediate cutoff. For developers, OpenAI models remain fundamentally accessible via API key, Codex, and compatible gateways.

The real task for teams is therefore not to prematurely switch editors, but to test their dependencies. Those who know today which workflows are tied to Cursor's own infrastructure and which can be mapped via independent model access can react to the final contract termination without having to rebuild their development processes on short notice.