Free and open source

See what your AI coding agents really cost.

One command reads the usage logs Claude Code, Codex, Copilot CLI and other agents already keep on your machine, and prices every model at API rates. See what your Claude plan is really worth. No signup to try it, and your usage never leaves your machine.

Connect an account when you want more: a dashboard by agent and model, a monthly plan-value receipt, and a place in the shared picture of what AI actually costs.

npx mytokentracker

Needs Node.js 20 or later. Free, with no paid plans. Your data is yours.

7 developers tracked their AI usage here in the last 7 days 60.2B tokens in the shared community data, last 30 days

Coding agent spend this month, at API prices

$184.20

Claude Code

claude-opus-4-5

$92.40

Claude Code

claude-sonnet-4-5

$48.10

Codex

gpt-5-codex

$28.55

Copilot CLI

claude-haiku-4-5

$15.15

Illustrative sample data. Yours comes from the usage logs your agents already keep.

Coding agents sync with one command. Your own apps take one SDK line or one API call.

Why it exists

AI got expensive, and nobody can see where it went.

You are paying for Claude, ChatGPT, and raw API calls, each in its own dashboard, none of them talking to each other. MyTokenTracker is the one place that answers a simple question: what is AI actually costing me, and on what?

1

Every coding agent, one view

Claude Code, Codex, Copilot CLI, Antigravity and more in one table, plus OpenAI, Anthropic, Gemini, Mistral and any other provider from your own apps through the events API or a drop-in SDK wrapper.

2

Is your plan worth it?

Your subscription hides what your usage would cost at API prices. The plan-value receipt puts the two side by side, so you know whether Max is paying for itself.

Run the numbers
3

Stay ahead of limits and hikes

Watch your usage against plan limits, and track a price-hike radar so the next increase is never a surprise.

Leaderboard

See who is shipping the most AI

A public leaderboard of real AI usage. Rank by tokens or spend, climb a streak, and share a profile card that shows your tokens, cost, favorite model, and a full activity heatmap. Opt in anytime, it is free.

1 113.2B$85K
2 146.2M$438
3 52.5M$157

A taste. The real board is live and growing.

Ship it fast Next.js

Ship it fast

Next.js · Stripe

AI Resume React

AI Resume

React · OpenAI

PriceSpy Laravel

PriceSpy

Laravel · Claude

VoiceNote Python

VoiceNote

Python · Whisper

Preview images auto-fetched — no upload needed.

Builder Showcase

Show what you ship. Get found.

Your public builder profile shows your real AI usage alongside the projects you ship. Add a URL and we automatically grab a preview image — no screenshot, no upload. Then your work is discoverable in our growing community showcase.

  • ✓ Public profile at mytokentracker.io/u/you
  • ✓ Up to 12 projects with auto-fetched thumbnails
  • ✓ Verified AI usage stats — not self-reported
  • ✓ Activity heatmap, skills, social links — all free

How it works

One command. Then forget about it.

  1. 1

    Try it, no account

    npx mytokentracker prints what your coding agents cost over the last 30 days, by day, agent and model. Nothing is installed and nothing is uploaded.

  2. 2

    Connect an account

    npx mytokentracker init uploads the history your agents still hold, then syncs every 30 minutes in the background (launchd on macOS, cron on Linux, a Task Scheduler job on Windows).

  3. 3

    See exactly what it costs

    Your dashboard fills with spend by agent and model, with your plan-value receipt, weekly trends and budgets. For your own apps, add a Python or Node wrapper or post to the events API.

Read the docs

# See what your coding agents cost (no account)

npx mytokentracker

# Is your $200 Max plan paying for itself?

npx mytokentracker --plan 200

# Track it over time (with an account)

npx mytokentracker init

# Any app, Python

track_openai(client.chat.completions.create(...))

Get started

Sign in with GitHub, Google, Microsoft or a magic link. No password.

We are a team

Together, we are mapping what AI really costs.

No one truly knows what AI costs in the real world, across tools, models, and the way people actually work. So we are figuring it out together. Your usage joins an anonymized, aggregate picture, and the more of us who track, the clearer that picture gets for everyone.

Anonymized and aggregate only. No names, no projects, no individual numbers. You can opt out anytime.

Explore the community dashboard
3,989
Models priced, updated daily
Anonymous
Aggregate only, opt out anytime
Live
Refreshes as the community tracks
$0
Free, to contribute or just look

Questions, answered.

How do I see what Claude Code costs me?+

Run npx mytokentracker in a terminal. It reads the usage logs Claude Code already keeps on your machine and prints the last 30 days by day, agent and model, priced at API rates. No account, nothing to install, and your usage never leaves your machine.

Which coding agents does it support?+

Claude Code, Codex, GitHub Copilot CLI, Antigravity, OpenCode, Amp, Qwen Code and more. The CLI is built on the open source ccusage, so it reads every agent ccusage reads. For your own apps, the events API and drop-in Python and Node wrappers track OpenAI, Anthropic, Gemini and any other provider.

Is my Claude Max plan worth it?+

Run npx mytokentracker --plan 200 (or 100, or 20) and it compares what your usage would have cost at API prices with what you pay for the plan. With an account, the plan-value receipt does this for you every month and makes a shareable card.

What does it upload, and when?+

Nothing until you run npx mytokentracker init with an account. After that it syncs daily totals per agent and model every 30 minutes: token counts and cost. Never prompts, code, file names or project contents.

Is it really free?+

Yes. Free, forever. There are no paid plans. The CLI is open source on GitHub and npm.

How is cost calculated?+

At current API list prices from our daily-synced catalog of 3,989 priced models. If you send a real cost through the events API we store it as actual, otherwise we label it estimated, so a guess never reads as a fact.

Can an AI agent set it up for me?+

Yes. Ask your coding agent to run npx mytokentracker, or point it at mytokentracker.io/llms.txt and it will find the CLI, the events API and the wrappers.

Is my usage public?+

Only if you leave the leaderboard on. Synced accounts show aggregate stats and your GitHub profile (name, handle, avatar) on the public leaderboard by default. We never show prompts, project names or file paths, and you can turn it off anytime in Settings.

Guides: Is Claude Max worth it? · Claude Code cost · Codex cost · vs ccusage

From the blog

Fresh research on what AI costs

All posts

Free weekly digest

Get the weekly AI Cost Index

One email a week: what AI cost this week, the biggest price moves, and where it is all heading. Free, no account needed. Unsubscribe in one click.

No spam, no account. One click to leave.

Start seeing what your agents cost.

One command, no signup, nothing leaves your machine. Connect an account when you want the dashboard and the receipt.