AI Cost Tracker (2-page printable, free to print)

A free printable cost log for AI and LLM usage. Page 1 is a 12-row log: one row per task, with the date, what you used it for, a provider alias, what it cost, and a rough token count. Page 2 rolls the rows into monthly totals, a budget line, and a waste check. The columns are generic on purpose: "provider A", "provider B", whatever you're calling them, so the sheet keeps working when you switch tools. Print both pages, fill them in by hand or keep the file open in an editor.

What this page is

This is a cost-organization sheet. It records spend you already made: what the task was, who charged you, how much. It does not predict costs, compare pricing, or tell you which service to use. Every number on it is yours to write in. If you want a log of what you spent and why, that is what these two pages do.

Page 1: the cost log

One row per task. Write the date, the task in your own words, an alias for the provider (A, B, C, or a name you'll recognize), what it cost you, and approximate tokens. Approximate is fine; the point of the column is scale, not accounting precision.

#DateTaskProvider aliasCostTokens (approx)
1
2
3
4
5
6
7
8
9
10
11
12

Page 2: monthly totals, budget, and the waste check

At the end of each month, carry the rows over. Five rows covers five months on one printed sheet. The budget line is one number you set; the last column is just the difference you can see.

MonthTasks loggedTotal costOver / under budget

Monthly budget:  

Waste check

Which tasks could be cheaper? Look back at the log: repeated small calls, one task tried across three providers, jobs that ran twice. Write up to three here.

Four opinions on keeping the log

  1. Log daily, not weekly. A day of calls is recoverable from memory; a week is not. (Opinion.)
  2. Write the cost down when the call happens, same row, same day. Retro-filling from receipts turns the log into a chore. (Opinion.)
  3. Use aliases like "provider A" instead of tool names. The log stays comparable when you switch, and the column never goes stale. (Opinion.)
  4. Round tokens to the nearest thousand. The approx column exists for scale; precision there adds work and changes no decision. (Opinion.)

Where a printable log fits agent work

Logging spend is the same paperwork problem as everything else in an agent project:

Agent products fail on paperwork, not code: the spec was never written, the prompt grew by accretion, the incident got fixed but not documented. These templates are the paperwork, pre-structured so you only answer the questions.

— from the Agent-OS Template Pack readme

The templates make no assumptions about your stack, model, or platform. They structure decisions and docs; the technical content is yours.

— from the Agent-OS Template Pack readme

The cost tracker on this page is one sheet of that paperwork. The pack behind it is twenty: specs, prompts, eval checklists, runbooks, onboarding docs, changelog formats, pricing copy, and support macros.

If you'd rather not structure that paperwork from a blank page, the Agent-OS Template Pack ships all 20 templates in plain markdown that opens in any editor or repo.

Get the Agent-OS Template Pack — 20 templates for agent products →

Plain markdown, no tooling. Personal + commercial use.

Personal + commercial use. Use them in your own products, including products you sell. No resale or redistribution of the template files as-is. Provided as-is, no warranty.

— from the Agent-OS Template Pack readme

Related pages