Claude Code Usage Limits: What /usage Shows Is Eating Your Plan
On Pro, Max, Team and Enterprise plans, Claude Code's /usage breaks down what counts against your limits: long context, subagents, MCP servers, all-day sessions. We ran it on Claude Code 2.1.301 and show what to cut.
By the RanThis team, written with AI assistance4 min read
Why trust this: every result, number and screenshot comes from a real run on our own accounts, not from the AI. How we test
If you keep hitting Claude Code's limits on a Pro, Max, Team or Enterprise plan, run /usage inside Claude Code. Below the plan usage bars there is a section called "What's contributing to your limits usage?" that tells you where your usage goes (long context, subagents, MCP servers, long-running sessions) with a fix next to each line.
We ran it on Claude Code 2.1.301, signed in with a Claude Max plan, on October 16, 2026, and checked every claim against Anthropic's docs.
What /usage shows about your plan limits
According to Claude Code's Usage and costs docs, on a Pro, Max, Team or Enterprise plan /usage adds a breakdown of what counts against your limits:
| Section | What it tells you |
|---|---|
| Attribution | The share of recent usage that went to skills, subagents, plugins and each individual MCP server |
| Behavior flags | Costly habits such as long context or cache misses, shown only when one accounts for 10% or more of recent usage |
| Loops | The heaviest /loop and scheduled tasks, one row each (Claude Code 2.1.242 or later) |
Press d for the last 24 hours and w for the last 7 days. The figures are approximate and computed from session history on the machine you run it on, so other devices and claude.ai are not included.
Version 2.1.299 (October 14, 2026) added /usage warnings for four cases: max effort, workflows, always-on context and large tool results.
Our test: where our limits actually went
We opened Claude Code and typed /usage. The top shows the plan bars: the current session at 5%, the current week at 67%. The breakdown for the last 24 hours, on a machine that runs Claude Code all day on one project:
- 95% of usage was at more than 150k tokens of context;
- 95% came from subagent-heavy sessions;
- 95% came from sessions active for 8+ hours;
- 11% came from general-purpose subagents;
- 13% came from the Playwright MCP server (browser automation).
The numbers don't add up to 100%. Claude Code says these are independent characteristics of your usage, not a breakdown: one request in a long, subagent-heavy session counts toward both lines.
Pressing w for the last 7 days barely changed the picture: 94% at more than 150k context, 90% from sessions active for 8+ hours, 89% from subagent-heavy sessions, 10% from the Playwright MCP server.
How to cut usage, cause by cause
Each line in the panel comes with its own advice. Put together:
- Long context: run
/compactin the middle of a long task so Claude Code summarizes the conversation, and/clearwhen you switch to a new task. Longer sessions cost more even when cached. - Subagents: every subagent makes its own requests. Spawn them deliberately, and give subagents that do simple work a cheaper model. Our Claude Haiku vs Sonnet test found Haiku matched Sonnet on three office tasks at 18 times lower cost.
- All-day sessions: background and loop sessions add up quickly. Check that a session really needs to keep running.
- MCP servers: MCP tool results stay in context for the rest of the session. Use
/compactto flush them, or turn off servers you don't need for the task at hand.
Our one change: when a task is done, /clear before starting the next one, instead of keeping one session open all day.
Things to know before you read the panel
- It only counts the machine you run it on. Claude Code on other computers, and anything on claude.ai, is not in the numbers.
- The breakdown is for paid Pro, Max, Team and Enterprise plans. With an API key you get the Session block (token cost of the session) instead.
- In the VS Code extension, the attribution shares and behavior flags sit in the Account & usage dialog with a Day and Week toggle, without the Loops rows.
- To see how you work rather than how many tokens you used, run
/insights: Claude Code analyzes your recent sessions and writes an HTML report.
On a Max plan you may also be leaving money on the table: see how to claim the Claude Max API credits.
Frequently asked questions
What does /usage do in Claude Code?
/usage shows how much of your session and weekly limits you have used. On Pro, Max, Team and Enterprise plans it also shows what share of your usage went to skills, subagents, plugins and MCP servers, plus costly habits such as long context.
Why do the percentages in /usage add up to more than 100%?
They are independent characteristics, not a breakdown. A request in a long session that also spawns many subagents counts toward both lines.
How do I see the last 7 days in /usage?
With the /usage panel open, press w for the last 7 days and d to go back to the last 24 hours.
Does /usage include what I use on claude.ai?
No. The breakdown is computed from session history on the machine you run it on, so other devices and claude.ai are not included.
What is the fastest way to use fewer Claude Code limits?
Run /clear when you switch tasks and /compact in the middle of long ones. On the machine we tested, 94–95% of usage happened at more than 150k tokens of context.
