How to Check Kilo Code Usage and Cost
Kilo Code prints a dollar figure for the task you just ran — but nothing that adds up across a week. Here is how to get the running total.
Quick answer
npx whoburnedmore— it reads Kilo Code's local logs, totals tokens by day and model, estimates cost, and ranks you. Kilo shows you the price of a single task in the editor, but it keeps no cross-time tally, so the CLI is how you see the bigger number. 🔥Kilo Code is an open-source coding agent for VS Code, JetBrains, and the terminal. It started as a superset of Cline and Roo Code, folding the best of both into one extension, and it became the natural place to land for anyone whose Roo Code workflow was orphaned when that project was archived upstream. If you searched for “Roo Code usage” and ended up here, that migration is exactly why — the tool moved, the question did not.
How do I see my Kilo Code usage?
The in-editor panel is great for the task in front of you and useless for the month behind you. To pull the historical picture, point whoburnedmore at the logs Kilo already writes locally:
$ npx whoburnedmore↳ reading Kilo Code local logs…↳ parsed 312 task records across 4 workspaces KILO CODE — BY DAY ──────────────────────────────── 2026-06-17 in 1,902,400 out 121,800 2026-06-16 in 2,640,900 out 198,300 2026-06-15 in 744,100 out 52,700 BY MODEL (via your provider) anthropic/claude-sonnet 9.4M google/gemini-flash 3.1M 7-day total: 12.5M tokens · est. at your rate
Because Kilo is bring-your-own-key, the token counts are real but the price tag depends on whichever provider sits behind your key. whoburnedmore sums the raw counts first and applies your model's published rate second, so the dollar figure tracks what you are actually billed rather than a guess.
The three numbers Kilo gives you in-editor
- 1
Per-task cost — shown the moment a task ends
Each completed task prints what it spent. Precise, immediate, and gone the next time you start something new — there is no “sum these up” button. - 2
Context fill — how much of the window is in use
A live gauge of the current conversation, not a usage total. It tells you when the agent is about to truncate, not what you have burned this week. - 3
Provider dashboard — if you use OpenRouter
Your key's billing page has spend, but it mixes every app on that key and is not broken out by Kilo workspace, day, or model the way a usage audit wants.
What does Kilo Code cost per task?
Kilo itself is free and open source — the spend is entirely the model calls behind your key. A single task's cost is just its input and output tokens priced at your provider's rate 💸:
The catch is the multiplier hiding in agent loops. An “edit, run tests, read the failure, edit again” cycle resends the growing context on every turn, so input tokens stack up far faster than the output you can see. That is why a task that felt small can still post a surprising number — and why the running total matters more than any single figure.
Bring-your-own-key means rates are yours, not ours
whoburnedmore does not invent Kilo Code pricing, because there isn't one. It reads the model names from your logs and applies the public per-token rate for each one. Swap to a cheaper model mid-week and the estimate follows it task by task.How do I track Kilo usage over time?
Over-time tracking is the gap the editor leaves wide open. whoburnedmore turns the pile of per-task records into a timeline, so a quiet research afternoon and a runaway refactor sit next to each other and the spike is obvious. The pipeline is short:
Kilo logs
local task records
npx whoburnedmore
parse + total
by day / model
with cost estimate
your rank
on the board
Day-level catches the runaway sessions. Model-level answers whether your default pick is worth its premium or whether a cheaper model would do. And if you also drive Claude Code, Codex, Gemini CLI, or Copilot, the same command stacks Kilo beside them — see the cross-tool usage guide for that combined view.
| Where you look | Single task | Day-over-day | Per-model split | Other agents |
|---|---|---|---|---|
| Kilo in-editor cost | — | — | — | |
| Provider billing page | aggregated | — | — | |
| whoburnedmore | 12+ |
Does the CLI touch my API key or code?
No. whoburnedmore reads only the numeric usage fields from Kilo's local records — token counts, model names, timestamps. It never opens your API key, your prompts, or your source files, and nothing is uploaded unless you choose to submit the aggregated totals to the leaderboard. Run npx whoburnedmore --dry-run to print exactly what a submission would contain before anything leaves your machine. 🛡️
Coming from Roo Code?
Your old Roo workflow maps straight across — Kilo is the superset, so the modes and the keybindings you knew still apply, and your usage question has the same answer. Point the CLI at the new logs and the history picks right up where Roo left off.command, no install
key — your provider rate
of history in one pass
Related guides
How to Check Your AI Coding Token Usage
The cross-tool overview: one command that totals your token usage and cost across every AI coding agent you run.
How to Check opencode Token Usage and Cost
opencode keeps detailed logs — here's how to turn them into a token + cost total.
A ccusage Alternative With a Leaderboard
Built on the same data as ccusage — with a cross-tool dashboard and a leaderboard on top.