Hey Cursor team and community,
I’ve been a loyal Pro user for over a year and Cursor has genuinely transformed how I code. But the removal of truly unlimited Auto (which was one of the biggest reasons I chose Cursor) has made daily work noticeably more expensive and stressful.
Auto used to be the reliable “just works” fallback for routine tasks, tab, debugging, and light agent work. Now it’s metered at API rates and I’m hitting limits way faster than before, especially when combining with Cloud Agents. The $20 included usage disappears in a few heavy days, and overage fees add up quickly.
Two concrete suggestions that would make Cursor way more competitive and user-friendly:
Unlimited access to a fast & cheap model like GPT-5 Mini (or equivalent)
GitHub Copilot gives essentially unlimited usage of their capable models for just $10/month. Why can’t we have something similar in Cursor? Let Pro users have unlimited GPT-5 Mini / fast cheap model for everyday coding, autocomplete, and simple agents. Save the expensive frontier models and limited quota for complex tasks only. This single change would instantly make Cursor feel affordable again for most developers.
Let Cloud Agents connect to our own VMs (self-hosted / BYO compute)
Cloud Agents are amazing because they have their own sandbox VM to build, test and iterate. But running everything on Cursor’s infrastructure is simply too expensive for frequent or long-running work. Please add the option to point agents to our own VMs (AWS, GCP, Azure, or even local with proper setup). There is already a feature request for self-hosted agents – this would be a massive win for power users and companies who want cost control.
I really don’t want to leave Cursor — the UX and agent features are still the best in the market. But if these pain points aren’t addressed soon, I (and quite a few others I’ve talked to) will have no choice but to switch to GitHub Copilot + VS Code or other more affordable alternatives.
What do you guys think?
How are you handling the new Auto limits?
Would unlimited GPT-5 Mini + self-hosted agents make you stay?
Cursor team — any plans or timeline for these kinds of improvements? Looking forward to the discussion!