Claude Code usage, limits and token cost: a guide
On this page
Run /usage in a Claude Code session to see your plan limits and which skills, subagents, plugins and MCP servers used them. Everything else on this page is a shorter route to the question you actually have.
How do I check my token usage?
/usage for the official limit and a 24-hour or 7-day attribution, /context for what fills the window right now, a status line for a number that stays on screen, ccusage for history from the command line, and a menu bar monitor for a live view without opening a session. Each shows something different.
Full guide: How to check Claude Code token usage: 5 ways.
How do the 5-hour and weekly limits work?
On a subscription, usage counts against a rolling five-hour session window and a weekly cap at the same time, and the two run out at different speeds. The post explains how to see what is left and what to do when you hit one.
Full guide: Claude Code 5-hour limit explained.
Why do my tokens run out so fast?
Usually one of a handful of causes: long context, cache misses, MCP results, memory files, subagents or loops. The fix depends on which one it is, so measure first.
Full guide: Why Claude Code burns through tokens and how to find the cause.
What do MCP servers cost?
A server adds its tool names and instructions to the context at session start; tool definitions are deferred by default so only names load until a tool is used. Large tool results are what get expensive.
Full guide: Claude Code MCP token usage.
Which monitor should I use?
- Mac, with a breakdown by source: Claude Code usage monitor for Mac.
- Windows or Linux: Claude Code usage monitor for Windows and Linux.
- Comparing options: SkillKeeper vs ccusage, vs the built-in /usage, vs other menu bar monitors.
Checked against the Claude Code docs on 2026-10-14. Limits and commands change between versions, so each post carries its own date.