Claude usage is the share of your plan's allowance that your chats, Claude Code sessions and Desktop work have used so far, measured against a five-hour session limit and a weekly limit. You can read it in Settings > Usage on claude.ai or with /usage in Claude Code, and keep a longer record in UsageAtlas, a free desktop tracker.
Everything below comes from Anthropic's help center and the Claude Code docs, checked on November 3, 2026. Anthropic reserves the right to adjust limits at its discretion, so treat the details as a snapshot.
One pool for every Claude surface
Anthropic counts everything you do with Claude against the same allowance. A long chat on claude.ai, an agent run in the terminal, and Claude Code inside VS Code or a JetBrains IDE all draw from one pool. Heavy chatting in the afternoon leaves less room for coding in the evening.
There is no fixed message count. How fast the pool drains depends on:
- how long your messages, attached files and the conversation so far are
- which model you pick and the effort level you set
- tools such as web search, Research and connected apps
- multi-step work like running code, creating files and browsing
The two windows
| Window | When it starts | When it refills | Who has it |
|---|---|---|---|
| Session | With your first message | In full, five hours later | Most plans; verified Ultra accounts are exempt |
| Weekly | At a time fixed for your account | The same day and time every week | All paid plans |
The weekly reset does not follow your billing date or the moment you start working. Anthropic assigns each account its own time, and Settings > Usage shows it. Max and Ultra also carry a separate weekly limit for Fable models.
Both windows fill at once. Anthropic's error reference warns that a single heavy burst, such as a large workflow fanout, can empty the weekly allowance before the session window has refilled.
Where to read your Claude usage
- Settings > Usage on the web or in Claude Desktop: a Current session bar with the time left, weekly bars with the next reset, and the usage credits section.
/usagein Claude Code: plan bars, activity stats and, on subscription plans, a breakdown of what used the most, by skill, subagent, plugin and MCP server./costand/statsopen the same screen.- Warnings: before a window runs out, Claude Code prints how much of it you have used and when it resets.
The /usage breakdown is built from local files on that one computer, so claude.ai and your other machines are missing from it. Its dollar figure is a list-price estimate, and on Pro, Max or Ultra it is not what you pay. Claude Code also deletes local session files after 30 days by default.
Keep a record with UsageAtlas
UsageAtlas covers what those screens leave out. It runs on Windows, macOS and Linux, uses the Claude sign-in already on your computer, and shows the five-hour and weekly windows as remaining capacity with local reset times. Pick a threshold for each window and it sends a desktop alert before you run low.
It also stores Claude Code tokens, requests, models and estimated cost by day, so last month survives Claude Code's cleanup, and it puts Codex, Cursor and other coding tools in the same view. It is free and open source, needs no account for local tracking, and does not read prompts or responses. Its cost numbers are estimates, not a bill.
If you only chat in the browser, Settings > Usage tells you all you need. A tracker earns its place once Claude Code is part of your working day.
When the pool runs dry
You can wait for the reset time shown in the limit message, turn on usage credits to keep going at standard API rates under a monthly cap you set, or move to a plan with more room. As of November 2026, Max ($100 a month) gives five times Pro's usage every five hours and every week, and Ultra ($200 a month) gives ten times Pro's weekly usage.