/usage in Claude Code: which skill and which MCP are eating your limit

TL;DR Don't open the screen, run claude -p "/usage" instead: it prints both windows at once. Then act on the name in the 7-day block, not the 24-hour one, because the cap that locks you out for days is the weekly one.

Update, October 31, 2026: Since v2.1.305 the block is titled What's contributing to your usage? and each warning is a one-line fix that names its source, such as Turn off wmedia when you don't need it. On the screen, f applies the fixes it can, and the same fixes show up in VS Code and /doctor.

When you run out of quota, the usual suspect is the model. Opus costs more, sessions get long, tough luck. That is where most people stop, because until recently there was nothing else to look at.

/usage now carries a block titled What's contributing to your usage? that splits your recent consumption across things with actual names: your skills, your subagents, your plugins and your MCP servers, each with a percentage. Above those lists it puts the fixes, one row each, with the share of your usage the fix applies to: turn off an MCP server, move a subagent to a cheaper model, start fresh when you switch tasks. A row only shows once that share reaches 10%. The name lists skip that filter entirely: a skill shows up at 1%. Prompt suggestions, the gray text after each reply, are a background request that counts toward your limit too.

This is how it looked on my machine in August, untouched, back when the warnings were behaviors rather than fixes:

$ claude -p "/usage"

Current week (all models): 11% used · resets Aug 20 at 8pm (Europe/Madrid)

What's contributing to your limits usage?
Approximate, based on local sessions on this machine — does not include other devices or claude.ai. Behaviors are independent characteristics, not a breakdown.

Last 24h · 155 requests · 3 sessions
  91% of your usage was at >150k context
  Top skills: /linkedin-daily 17%, /claude-code-tip 12%
  Top MCP servers: wmedia 19%

Last 7d · 2699 requests · 10 sessions
  88% of your usage came from sessions active for 8+ hours
  83% of your usage was at >150k context
  54% of your usage came from subagent-heavy sessions
  Top skills: /linkedin-daily 3%, /claude-code-tip 2%, /evaluate 1%
  Top subagents: general-purpose 5%
  Top MCP servers: plugin:chrome-devtools-mcp:chrome-devtools 13%, wmedia 3%

And look at what the two windows do: they disagree.

Over the last 24 hours the heaviest MCP server is wmedia at 19%, and my publishing skill takes 17%. Stop reading there and I disconnect wmedia and call it a day. But across 7 days wmedia drops to 3% and the top spot goes to plugin:chrome-devtools-mcp:chrome-devtools at 13%, a server I never configured by hand. A plugin brought it in, which is why the name starts with plugin:.

One atypical day is all it takes to convict the wrong suspect. Claude Code knows it too: when nothing in a window clears 10%, it sends you to the other one.

Nothing over 10% in this period — try the other window.

Since v2.1.305 the block reads differently, but the trap is the same. This is the screen today, in the cloud environment where my routines run:

What's contributing to your usage?
Approximate · sessions on this machine only

Last 24h · shares overlap

100% Give "general-purpose" subagents a cheaper model
40%  Start fresh when you switch tasks · run /clear
21%  Turn off wmedia when you don't need it

Subagents               % of usage
general-purpose                93%

MCP servers             % of usage
wmedia                         21%
Notion                          1%

f to fix 2 of 3 · d to day · w to week

It no longer tells you that 21% came from wmedia: it tells you to turn it off when you don't need it. If that 21% comes from an atypical day, so does the fix.

Why the week beats the day

You have two caps. The five-hour window, which you can wait out over a coffee, and the weekly one, which locks you out until its reset date. The second is the one that hurts. On Ultra with a verified account only the weekly one is left.

The interactive /usage panel always opens on the last 24 hours. This is not a saved preference or something your config decides: it starts on the day, every time, and you have to press w for the week. So the instrument it hands you by default is calibrated for the wrong problem.

One detail worth keeping straight: Last 7d is a rolling seven days, not your billing week. Your weekly bar resets on a fixed date. The two periods look alike and are not the same span.

How to use it

1. Ask for both windows at once

claude -p "/usage"

No TUI, no keystrokes. It prints the 24 hours and the 7 days back to back, each row with its percentage, and adds the request and session count for each period, which the interactive screen leaves out.

What this route does not print, if you are on a subscription, is the Session block. That is where the Prompt cache (main) line lives, the one that tells you whether your cache is working and what broke it. For that one you have to open the screen inside the session.

2. Read the 7-day block only

Skip the 24-hour one. Find the first row that names something of yours: a skill, a subagent, a plugin or an MCP server. That is your suspect.

3. Apply the fix with f

Every row already is the fix, so there is nothing to translate. Open /usage inside the session, switch to the 7-day window with w, and press f when the footer says f to fix 2 of 3 or f to fix both. Claude Code runs the rows that end in an /effort command itself, and for the others it knows how to handle it hands Claude a single prompt that shows you each change before making it. The ones that end in run you run yourself: Compact before you step away · run /compact, or the /clear when you switch tasks on the long-context row. The same fixes show up in the VS Code Account & usage dialog and in /doctor.

What it can name

What these numbers are not

They are not an audited bill. The block says so itself: the figures come from local sessions on that machine, so anything you did on another device or on claude.ai is missing. One request can count toward several rows, which is what shares overlap means, so they can add up past a hundred between them. And each row's percentage is the share of your usage the fix applies to, not what it saves.

A version warning too: before v2.1.222, a single call to an MCP server was enough for Claude Code to attribute every later request to it, inflating its share. If you looked at this screen a few weeks ago and blamed a server, look again. And since v2.1.299 it flags four more causes, each with its fix.

There is also one angle this screen cannot cover: Top skills only names skills that actually ran. The ones you never invoked consume nothing, so they never appear here, and they still take up their line in the prompt on every turn. For those, there is /skill-doctor.

And if what you get is Compact before you step away, the cache-miss row, this screen tells you it happened but not what caused it or when. That comes from the Prompt cache (main) line in /cost, which names the cause of your last miss.

For the rest of the picture: /usage and /stats as gauges tell you how much is left, the status line keeps it in front of you without typing anything, and if context is what got flagged, connecting twenty MCP servers without eating your context and the auto-compact window are the two levers left. And if what got flagged is subagents, like the 54% of subagent-heavy sessions in my August capture, the lever is sizing your workflows before you launch them.

Official docs: Manage costs effectively

Requirements: the one-line fixes and the f key need Claude Code v2.1.305 or later, on a Pro, Max, Ultra, Team or Enterprise plan.