Last week, OpenAI announced they are cutting off Cursor’s native access to its models on November 12. The reason: SpaceX now owns Cursor, and OpenAI can’t be confident SpaceX will use its technology within OpenAI’s terms of service. Developers who built workflows around OpenAI models inside Cursor now have to scramble: bring their own API key, install a separate extension, or find another way in.
Nobody using Cursor did anything wrong. They built on a tool where the model access ran through someone else’s contract, and that contract changed hands.
The question is who controls your model access #
Ask any developer what they want from an AI coding tool, and you’ll hear some version of the same thing: the freedom to use the best model for the job, without waiting on permission. Ask their leadership what they want, and you’ll hear a different list: visibility into what’s running, what it costs, and whether it’s compliant. Both are describing the same gap.
Most companies end up choosing one of two flawed paths. Lock everything down to a single tool and a single model provider, and your best builders spend their time working around the constraints instead of building. Or look the other way while developers use whatever they want, with no governance and no visibility. Neither is a robust, long-term strategy.
Kilo was built model-agnostic from day one #
Kilo is open-source, and it connects to more than 500 models across 60-plus providers. Developers route each task to whichever model fits: cost, capability, or latency. There’s no proprietary gateway sitting between a developer and the model they want to use, and no single vendor whose business decisions can take the tool down.
It’s the design principle upon which Kilo grew from zero to 3 million developers, driven almost entirely by word of mouth.
Why this matters even more now #
The Cursor situation is a clear, public example of what happens when model access depends on a commercial relationship. A change of ownership, a licensing dispute, a strategic pivot on someone else’s side of the table, and any of it can interrupt your team’s ability to ship. Cursor users got eleven weeks’ notice. That is not enough time to re-architect how a team of 200 developers writes code.
Model agnostic isn’t a nice-to-have feature. It’s insurance against exactly this kind of disruption; the freedom to keep building no matter what happens upstream.
Freedom and governance aren’t a trade-off #
Now part of Anaconda, Kilo connects to the governance layer: policy enforcement, provenance, cost visibility, applied the moment a developer writes the first prompt.
The platform underneath should be open and agnostic by default, so builders move fast, and leaders retain visibility and control.
Developers can get started with Kilo today. Or, explore Anaconda Platform for your organization. Contact us to discuss your AI needs.