This is not about Cursor Tab’s code autocompletion. This is about the “Agent Autocomplete” feature which autocompletes the prompt text as you type.
Does Cursor’s “Agent Autocomplete” feature consume tokens from my subscription? If it does, what model does it use and how can I know how much I’m spending by using it?
Thanks!
PD: Please excuse me if these are already answered in the official documentation, I couldn’t find the answers.
Hey, good news, Agent Autocomplete doesn’t spend tokens or count toward your subscription usage, it’s completely free.
Unlike Tab (code completions), it doesn’t call an AI model at all. The suggestions are generated locally on your machine using lightweight heuristics, like file names from your project, symbols from code you recently viewed, and words you’ve already typed in the prompt. Nothing is sent to the server, so there’s no cost, and it won’t show up in the usage dashboard since that only includes billable model requests.
If you want to turn it off, there’s a toggle called Agent Autocomplete in Cursor Settings > Chat.
And don’t worry if you didn’t see this in the docs, it’s really not the most obvious detail.