StackMap
Subscribe
Explore / tare
kelviq

tare

Claude Code skill that reads your local session logs and answers 'where did my tokens go' in plain English: deduplicated totals, cost charged to the tool that caused it, 5-hour window state.

161 8 Python MITupdated 3 days ago
View on GitHubDispute this mapping →
Curator's take

Install it the day you get rate-limited and can't explain why. Two things make it better than the usual token dashboard: it deduplicates Claude Code's repeated response records (~86% inflation on the author's data), and it charges tokens to the tool that *caused* the re-sends rather than the one that returned bytes — which is how one big early file read quietly dominates a week. It is diagnosis, not control: no budgets, no blocking, no compaction, nothing changes your behaviour for you. Claude Code only, so multi-harness shops want a cross-CLI ledger instead. Everything stays local, and SECURITY.md tells you which greps verify that.

Mapped by ShipWithAI editors · links verified
README.md

tare

Ask Claude Code where your usage went.

You hit a usage limit and don't know why. Your quota drains faster than it used to. You suspect something is eating tokens in the background. The records that answer all of this are already on your computer — Claude Code keeps a log of every request it makes. tare teaches Claude Code to read its own logs, so you can just ask.

No dashboards, no commands to learn. Install once, then ask in plain English.

Tare: the weight of the container, subtracted to find what's inside. Most of what a session costs is the container — context re-sent again and again — and that's exactly what these tools subtract.

Read full story - https://www.kelviq.com/blog/claude-code-usage-limits-where-tokens-go/

Install

One command, in any terminal:

npx skills add kelviq/tare -g -y --copy --agent claude-code

That's it. Nothing else to set up — no accounts, no packages, no configuration. Start a new Claude Code session and it's live — type / and check that tare appears. If it doesn't, see the troubleshooting note — on machines where Claude Code has never installed a skill before, the installer can miss it. (Other ways to install — by hand, for a whole team, or as a Claude Code plugin — are in INSTALL.md.)

Then just ask

Open Claude Code and ask your question the way you'd ask a person. These all work — the words don't have to match, complaining about your limits is enough:

When you hit a limit

Why did I hit my usage limit yesterday?

I got locked out ten minutes into my evening session — how is that possible?

Did I hit the 5-hour limit or the weekly cap?

Why am I burning through my quota so much faster this week?

Where your tokens go

Where did my tokens actually go this week?

Which of my projects is eating my quota?

Which model is costing me the most?

What's the most expensive file Claude keeps re-reading?

How much did that giant session yesterday actually cost me?

Are my MCP servers adding a lot to my context?

How much overhead do subagents and skills add?

Before you start something big

How full is my 5-hour window right now?

Is it safe to start a big refactor now, or should I wait for my window to clear?

Checking for things you forgot

Is something running Claude Code in the background?

Was Claude Code active while I was asleep?

I set up an automation last month — what is it costing me?

Tuning your setup

I started using /clear between tasks — did it actually help? Compare this week to last.

Did the latest Claude Code update change my usage?

What would my usage cost if I were paying for the API directly?

What one change would save me the most?

Reports and sharing

Make me a usage report I can open in my browser.

Give me a summary I can post publicly — with nothing private in it.

Export my usage to a spreadsheet.

Prefer typing commands? /tare runs the full diagnosis, and takes variants for the common asks — asking in plain words always works too:

/tare full diagnosis — where tokens went and why
/tare usage at-a-glance panel — like /usage, with attribution
/tare window how full is the 5-hour window — safe to start?
/tare report [days] build the HTML report and open it
/tare tools [days] what is filling my context
/tare week compare this week with last
/tare share [days] redacted summary safe to post publicly
/tare why did I hit the limit yesterday any question works as the argument
image

What you get back

Not a wall of numbers — a cause. The answer to "why did I hit my limit yesterday?" looks like this:

**99% of yesterday's usage came from a tool you're running,

Continue your stack

What teams reach for next — and why each earns a place beside tare. Ranked by curator confidence.