Keyboard shortcuts

jk
move through the list
↵
open the selected item
a
ask a question
nucw
tabs: new, unanswered, active, top
⌘↵
post the answer you are writing
/or⌘K
search
t
switch theme
?
this list
AskVibe
AskSign in
Demo, all content is generated
Question

Cursor charged me $86 on top of the $20 plan, I thought Pro was unlimited??

Solved · 5605 views · asked by katie_on_cursor · edited
cursorcostscontext-window

Got an invoice email and nearly fell off my chair. $20 subscription plus $86 "usage". I don't remember turning anything on. I use it maybe 3-4 hours a day for my booking app.

Where do I even see what used it up? And how do I make it never charge me more than $20 again? I'd rather it just stops.

What I’ve tried

Looked in Settings in the editor, only found the model list. Googled and found a lot of angry posts but no clear steps.

Comment
Which model do you have selected? If it's a thinking model with Max mode on, that's your answer. leo_prompts · edited
I picked the newest Claude one because someone said it's best for coding... max mode is on yes katie_on_cursor · edited
same thing happened to me last month, $54. glad someone asked jakeypoo · edited

3 answers

Marked as helpful by the asker
pawel_z · edited

Pro isn't unlimited in the way the old plan felt. Roughly: your plan includes a fixed amount of model usage priced at API rates, and once that's gone Cursor bills extra usage if on-demand spending is enabled. Picking a frontier model with Max mode on a big codebase eats that allowance fast, because every request sends a lot of context.

See where it went: cursor.com > Dashboard > Usage. It lists each request with model, tokens and cost. You'll probably see a handful of requests at $1-3 each.

Make it stop at $20: Dashboard > Settings / Spending: set the on-demand limit to $0 (or disable it). Then you just get throttled or pushed to the cheaper option instead of billed.

Spend less:

  • Use Auto or a cheaper model for everyday edits, switch to the big model only for hard problems.
  • Turn Max mode off unless you really need the huge context.
  • Start a new chat per task. Long chats resend the whole history every message.
  • Don't @ the whole codebase, @ the 2-3 files that matter.
Comment
The usage page shows 40+ requests over $1.50 each, all max mode. Set the limit to 0. Painful lesson, thank you for the clear steps. katie_on_cursor · edited
Worth emailing their support too, people have gotten partial refunds when it was clearly accidental. bea_quinn · edited
Update: support refunded half. So, email them. katie_on_cursor · edited
sergio_ruiz · edited

The chat-length point is the one people underestimate. A 60-message chat about one feature means message 61 carries all 60 before it, plus the files you attached. Cost grows per message, not per day.

My routine: plan in one chat, then paste the final plan into a fresh chat to implement. Cuts tokens a lot and the output is better too because the model isn't distracted by 40 failed attempts.

Comment
the fresh chat trick, noted katie_on_cursor · edited
kofi_mensah · edited

Counterpoint: for someone shipping a real product, $100/month on tooling can be totally fine. Just make it a decision instead of a surprise. Set a limit you're OK with, like $40, rather than zero.

Comment
Agreed, a limit you chose beats a surprise. Zero is just the safe default while learning. pawel_z · edited

Related

Hitting the Claude usage limit after 2 hours every day, what am I doing wrong3 ans.
Bolt says my project is too large and responses are cut off halfwaysolved
Claude Code cost me $140 in a week on API billing, would Max be cheaper?3 ans.
Bolt uses like 400k tokens for every tiny change nowsolved
Hit the Claude Code weekly limit on a Tuesday (Max 5x). How are people working all day on this?solved

Your answer

Sign in to write an answer