Track Gemini CLI token usage

Counted from disk, history included. Gemini CLI records tokens for every turn in its saved chats; burnlog reads those and nothing else.

read from local logs · free · open-source CLI · tokens only

Counting Gemini CLI in under a minute

01
Paste the prompt into your agent
Copy the setup prompt from burnlog.net and paste it into any coding agent. It installs the CLI, shows you your numbers first, and asks before anything is uploaded.
02
Or run it yourself
```
$ npx @sxnalabs/burnlog
```
burnlog finds every agent on the machine, reads its usage, and offers GitHub sign-in.
03
Keep it current
```
$ burnlog install
```
Installs a hook so future sessions sync themselves.

What burnlog reads

Gemini CLI stores each session under ~/.gemini/tmp (or $GEMINI_CLI_HOME) with a usage block per model turn: input, output, cached and thinking tokens. burnlog takes the counts and the model name. The conversation itself is never read into the upload.

Cached input is split out from fresh input, and thinking tokens count as output, so a Gemini number means the same thing as a Claude or Codex one on the board.

What burnlog can't see

Sessions only count if Gemini CLI saved them. Chats cleared with Gemini's own cleanup are gone before burnlog can see them.

Put Gemini CLI on the board

One command. It scans, shows you your numbers, and only then asks for an account.

$ npx @sxnalabs/burnlog

See the leaderboard →

Other agents

Claude Code · Codex · Cursor · aider · opencode