Burnmeter

Guide

How to Export Your Cursor Usage CSV (and Read It)

Where to find the usage export in Cursor in 2026, what each column means, and how to turn it into real dollar costs.

Prices last checked 2026-11-09. Independent comparison; not affiliated with any vendor named.

Where the export lives

On a legacy Cursor plan (Hobby, Pro, Pro+, Ultra, Teams Standard or Premium) open cursor.com/dashboard and go to the Usage or Spending tab. Pick a date range, then press Export CSV.

On The X Subscription (Free, Plus, Premium, Super, Ultra, Teams), usage lives under Usage at grok.com/dashboard, and your editor settings show the same remaining pool.

Export a full billing cycle if you can. Thirty days of events gives a far better plan recommendation than three.

What the columns mean

Date: when the request ran. Kind: whether it was included in your plan, billed on demand, or errored and not charged. Model: the model that served it, such as claude-opus-5.5-thinking, grok-4.7 or composer-2.5.

Input (w/ Cache Write) and Input (w/o Cache Write): the difference between the two is the tokens written to the prompt cache. Cache Read: tokens re-read from cache, usually the biggest number by far in agent sessions. Output: tokens the model wrote.

Cost: often "Included" or empty. That is why Burnmeter rebuilds the dollar amount from the token columns at each model’s list price.

Turn it into dollars

Drop the file on Burnmeter. It is parsed in your browser, priced row by row with the maintained price table, and run through every Cursor and Claude plan to show which one would have been cheapest.

Quick calculator

Estimate

About $0.93 per active hour, so $78.12/mo at list price.

  • Claude Pro Cheapest $20.00
  • Pay-as-you-go API $78.12
  • Cursor Plus Can't run this model$86.12
  • Claude Max $100
  • Cursor Premium $108
  • Cursor Super $178
  • Claude Ultra $200
  • Cursor Ultra $278

Assumes about 2.2M tokens per active hour, mostly cache reads, which is typical for agent sessions. Your real mix can be very different: drop your own export for exact numbers.

FAQ

Does the Cursor CSV include my code or prompts?

No. It has one row per request with the date, model, kind and token counts. Burnmeter reads it in your browser either way.

Why is the Cost column "Included"?

Requests covered by your plan’s pool show as included instead of a price. The token columns still tell you what the request would cost at list price.

Related

What does your AI coding cost?

Drop your Cursor CSV or ccusage output. It's read in your browser and takes under a minute.

Analyze my usage free