AI NewsProductivityAnnouncement
tokentab reads coding agent session logs to show what they cost
tokentab is a local Python CLI that parses Claude Code, Codex and Gemini CLI session logs and breaks the cost down by model, project and day. 588 stars in four days, MIT licensed.
Image: GitHub
Why it mattersMost teams cannot answer "which project is spending our agent budget" because the answer is spread across three tools with three separate dashboards, and two of them do not have one.
tokentab is a command line tool that reads the session logs Claude Code, Codex and Gemini CLI already write to disk, and adds up token usage and cost by model, by project and by day. The repository was created on 27 August 2026 and had 588 stars four days later, under an MIT licence.
It runs entirely on your machine. No account, no API key, nothing uploaded.
Where it looks
Claude Code ~/.claude/projects/**/*.jsonl
Codex ~/.codex/sessions/**/rollout-*.jsonl
Gemini CLI ~/.gemini/tmp/**/session-*.json
Cursor support exists in the code but is described as an unfinished placeholder. Tools you do not have installed are skipped, so you only see what you actually use.
Using it
python cli.py # last 7 days, everything
python cli.py -today # just today
python cli.py --provider claude # one tool only
python cli.py --project myapp # one project
python cli.py --json | jq . # machine readable
python cli.py -web # the same numbers in a browser, with charts
The web view runs on localhost:4747 and lays the figures out like a monthly statement. Colour is turned off automatically when the output is piped, so pasting a summary into a pull request does not carry escape codes with it.
Why this is worth ten minutes
Three separate agents, three separate log formats, and in most cases no shared reporting. If someone asks which repository is consuming the agent budget, the honest answer in most teams is that nobody knows, because the data sits in three folders in three formats and nobody has combined them.
This combines them, locally, from files that already exist. There is nothing to instrument and nothing to send anywhere.
The per-project breakdown is the number that changes behaviour. Total spend is a line in an invoice and people pay little attention to it. "This one repository is two thirds of it" leads to a discussion about why, and usually the answer is a specific loop or a specific habit that can be fixed.
What to check before you rely on it
The repository is days old and the cost figures depend on a price table inside the tool. Prices change, and an out-of-date table produces a wrong number that looks certain. Read where the rates come from before quoting a figure to anyone who cares about it.
It reads logs, so it can only see what the logs record. A session that a tool did not log is a session tokentab cannot count.
Source
- damejan80/tokentab on GitHub, MIT licence. Star and date figures read from the GitHub API on 31 August 2026
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.