Like many of you, I joined Cursor because it offered something rare: a smart, developer-focused IDE with unlimited access to its Auto model. That was a core promise — and a game-changer.
But now? Cursor has silently removed unlimited access to Auto. No clear announcement. No transparency. Even Pro users are now subject to quotas and token limits. This is a betrayal of trust.
Let’s be clear: we understand that premium models like GPT-4 or Claude can’t be unlimited. But Auto was Cursor’s own model, designed for daily use. It was supposed to be the reliable, unlimited fallback. Now it’s behind a quota wall — and there’s no free, unlimited alternative. Not even Composer 1.
We’re calling for a boycott of Cursor until they offer at least one AI model with full, unlimited token access.
What we demand:
A truly unlimited model (Auto, Composer 1, or any fallback)
Transparent communication about pricing and policy changes
Respect for paying users and the original value proposition
What we’re doing:
Canceling Pro subscriptions
Migrating to alternatives like Codeium, Continue.dev, or Aider
Spreading the word across forums, Discords, and dev communities
Cursor was built on the trust and enthusiasm of developers. If we let this slide, it sets a dangerous precedent for the future of AI tooling.
Let’s make our voices heard. No unlimited model, no support.
Correcting some misinformation here for others reading this post.
Auto isn’t a model, and has never been. It’s an automatic router for models:
“Enabling Auto allows Cursor to select the model best fit for the immediate task and with the highest reliability based on current demand. This feature can detect degraded output performance and automatically switch models to resolve it.”
You’re right — I misspoke earlier. I do understand that Auto is a routing mechanism, not a standalone model. My frustration isn’t about the technical definition, but about the removal of unlimited access to what was effectively the most reliable and consistent experience for many users.
Whether it’s a model or a router, Auto represented a dependable, unrestricted workflow — and taking that away without a fallback or clear communication is the real issue here.
It’s not just about that — the real issue is that there’s no free model available with unlimited token usage. That’s the core of the problem.
if the router can’t pick the right model for the task, then why not offer two distinct AUTO modes?
It makes zero sense that something as trivial as changing a CSS color triggers a call to GPT-4.5 Opus. For that kind of task, even GPT-2.5 would be overkill. If the issue is project size or complexity, then introduce an “AUTO Light” and “AUTO Pro” mode — where Light never touches the latest, most expensive models.
That would be a fair compromise. But right now, it just feels like forced overbilling disguised as smart routing.
We’re calling for a boycott of Cursor until they offer at least one AI model with full, unlimited token access.