Claude Code insights: what your usage is telling you
A total tells you what you spent. Insights tells you what changed, which model drove it, and whether your plan is still the right one. It is part of the free dashboard and fills in as soon as your usage syncs.
What is on the page
-
1
This week vs last week
Cost, tokens and sessions compared with the previous seven days, so a heavy week stands out while it is still happening.
-
2
Daily cost: plan vs API
A day-by-day chart of your plan's daily cost against what the same usage would cost at API prices. The gap between the lines is the value your plan delivered.
-
3
Cost by model (30 days)
Which models carried your spend. Moving routine work to a cheaper model usually starts here.
-
4
Project economics (30 days)
Cost, tokens and the dominant model per project, for usage that arrives with a project name through the events API or the wrappers.
-
5
This month vs last month
The longer view: whether your usage, and the value of your plan, is growing or shrinking.
Get it running
- Create an account and copy your API token.
- Run
npx mytokentracker initand paste the token at the hidden prompt. - Open Insights from the dashboard. It updates every time the CLI syncs, every 30 minutes on macOS and Linux.
Prefer to try it without an account first? npx mytokentracker prints the last 30 days in your terminal and uploads nothing.
Frequently asked questions
What are Claude Code insights?
Insights is a page in your free MyTokenTracker dashboard. It turns your synced Claude Code and other agent usage into comparisons: week over week, month over month, plan cost against API cost, cost by model and cost by project.
What do I need to see Insights?
An account and synced usage. Run npx mytokentracker init, paste your API token, and the CLI uploads your history. Insights fills in from there.
Is Insights a paid feature?
No. MyTokenTracker has no paid tiers. The CLI is MIT licensed and the dashboard, Insights and the weekly email are free.
How far back does it go?
As far back as your synced history. Claude Code keeps local logs for 30 days by default, so the first sync covers about a month, and from then on the CLI keeps adding to it.
Does it read my prompts or code?
No. Insights works from daily token totals and cost per agent and model. Prompts, responses, code and file paths are never uploaded.