Every token your team spends, accounted for.
TokenWork tracks LLM token usage per project and provider. Set a monthly budget, hear about it at 80%, and share one report with the whole team.
1.64M tokens since Oct 1, 41% more than at this point in September.
The jump on Oct 12 comes from the summarize step. It moved from claude-haiku-5 to claude-sonnet-5, and output tokens per call went from 310 to 690.
By model
-
gpt-6-mini820K $18.40 -
claude-sonnet-5490K $25.90 -
gemini-4-flash210K $2.30 -
claude-haiku-5120K $1.20
$ tokenwork budget set support-bot 2000000 2 ✓ Monthly budget set to 2,000,000 tokens. Alerts at 80% and 100%. Counted on your machine. Read it from anywhere.
The CLI records token counts where your code runs and sends them to your dashboard. Open it on a laptop or a phone and today’s numbers are already there.
Today
1.82M tokens
$41.20 · 12% more than yesterday
By provider
support-bot
Monthly budget
1.64M of 2M tokens
80% alert sent to admins
By model
-
gpt-6-mini820K -
claude-sonnet-5490K -
gemini-4-flash210K -
claude-haiku-5120K
Resets Nov 1
Alerts
- support-bot passed 80% Today, 09:12
- nightly-evals passed 100% Yesterday, 18:40
- checkout-api passed 50% Oct 14
- September report shared with the team Oct 1
One project, every provider.
See what one project spends at OpenAI, Anthropic and Google side by side. Input and output tokens are counted apart, and each model is priced at its own rate.
See what moved the bill.
The breakdown splits a month by project, model and day, and puts the biggest change on top. You know which call got expensive before you open a single log.
Your providers and your team, in one place.
Run tokenwork providers add openai and that provider’s usage starts showing up. Invite teammates as Admins or Members: everyone reads the same reports, and Admins manage the budgets.
Budgets that speak up at 80%.
Give a project a monthly token budget. TokenWork tells the team at 80% and again at 100%, so an overrun lands in your inbox before it lands on the invoice.
From teams who use it.
Notes from people who keep an eye on their spend.
A test suite was calling a large model on every run. TokenWork showed it on the first day, and the fix took ten minutes.
The 80% alert is the feature I didn’t know I needed. Finance stopped asking me about the invoice.
Setup was one command per provider. Now every project has a budget, and nobody argues about who spent what.
The CLI and the dashboard always show the same numbers, so the monthly report goes straight into our planning doc.
Start counting on the machine you already use.
Install the CLI with npm, a standalone binary or Docker. Run tokenwork login, add a provider, and tokenwork open takes you to the dashboard.
TokenWork counts tokens. It doesn’t need the text of your prompts or responses to do that.