ClaudeMax
Session · 5h62% left
Week · all models31% left
Week · Opus54% left
Extra usage credits12 of 100 used
Headroom · for Claude people
A Claude usage tracker
in the taskbar, not another tab.
A 5-hour clock, a weekly clock, and on Max a third meter scoped to one model family. Headroom reads what your account reports and keeps every one of them live in the Windows taskbar.
Windows 10 & 11 · signed auto-updates · no account needed
How Claude's usage limits actually work
Two clocks, one shared pool, no published numbers.
On the paid plans, Claude runs two clocks at once. A session limit resets every 5 hours; a weekly limit resets at a fixed time each week assigned to your account. Anthropic publishes no absolute numbers for either — plans are sold as multipliers (Pro is at least 5× Free per session, Max 5× and Max 20× multiply Pro), and what a multiplier is worth has shifted repeatedly since August 2025. So a Claude usage tracker has one honest option: read the numbers Anthropic itself reports, and skip the token-count guesswork.
What drains them is everything: message length, attachments, conversation history, model choice, effort level, web searches, artifacts. A Claude Code session hauling a big project context spends faster than an afternoon of short chats ever will. At zero you get a notice and three options — wait out the reset, upgrade, or pay as you go with usage credits.
The session window
Every plan has one, Free included, and it resets every 5 hours. Anthropic's own pages say exactly that and no more; the claim that the clock starts with your first message is widely reported but not officially documented. Since May 6, 2026, Claude Code's 5-hour limits are doubled on Pro, Max, Team, and seat-based Enterprise, and its peak-hour throttling is gone for Pro and Max.
The weekly window
Introduced August 28, 2025 for Pro and Max, after Anthropic pointed at “extreme usage by a small number of customers”. It resets at a fixed weekly time assigned to your account; that time never moves, whichever day you subscribed. Run it dry and paid usage locks until the reset, unless you upgrade or switch on usage credits. Usage-based Enterprise orgs skip fixed limits entirely — they're billed on consumption.
The scoped weekly (Max only)
Max plans carry a second weekly limit that covers a single model family. It launched Opus-scoped, but Anthropic's help pages currently disagree about which models it covers now, and the usage API names the scope per model, week to week. When it empties, that model goes dark until the weekly reset; everything else keeps working.
Usage credits
The overflow valve, available on Pro and Max. Once a limit is hit you can keep working at standard API rates, drawn from a prepaid balance with a monthly cap you set yourself. Credits don't raise the plan's included limits; the 5-hour clock keeps resetting underneath while you pay to bridge the gap.
What a Claude usage tracker should show
Each window above becomes a live meter next to the clock — new ones Anthropic ships appear on their own, tagged new.
Twin bars, or one big number
Idle, the tray icon settles into twin bars: tightest session on top, tightest week below. In active use it becomes one big number, percent left on whichever window bites first. 62 means keep going; 8 means commit your work.
Alerts with a cooldown
One notification at your threshold, one at critical, and six hours of quiet per quota between repeats. Escalations break through; stale data never notifies. When a window empties, the meter flips to a countdown until Claude is usable again.
The Anthropic API, separately
Subscription meters and API bills are different things, so Headroom keeps them in different gauges. Give the Anthropic API gauge an Admin key and a monthly budget, and your org's month-to-date spend becomes one more draining bar, reset on the 1st. It warns; it doesn't cut you off.
Asked every five hours
When does my Claude usage limit reset?
Session limits reset every 5 hours. Weekly limits reset once a week at a fixed time assigned to your account; it never moves, and there is no universal reset day, so the “Monday morning” answers you'll find on forums are someone else's Monday. Your exact countdown is in Settings → Usage on claude.ai and under /usage in Claude Code — Headroom keeps the same countdown next to the clock.
Do Claude Code and claude.ai share the same usage limit?
Yes. claude.ai, Claude Code, and Claude Desktop all draw from the same pool, for both the 5-hour session window and the weekly window. Heavy chatting in the afternoon leaves less for coding in the evening, and the other way around.
Why did I hit my Claude limit so fast?
Consumption isn't per message. Long conversations, attached files, bigger models, higher effort, web search, and artifacts all cost more, and a Claude Code session carrying a big project context burns fastest of all. On Max, certain frontier models also draw down a second, tighter weekly bucket. The limits themselves move, too — a holiday boost that expired on January 1, 2026 made a lot of allowances feel suddenly smaller.
What happens when I hit my Claude limit — can I keep working?
You get a limit-reached notice, and that window locks until it resets. From there you can wait, upgrade, or enable usage credits in Settings → Usage: pay-as-you-go continuation billed at standard API rates, with a monthly cap you choose. Credits exist on Pro as well as Max, and they don't raise the included limits; the 5-hour clock keeps resetting underneath.
Is there still a separate weekly Opus limit?
Max plans still carry a second weekly limit alongside the all-models one, but its scope has drifted since the Opus-only launch in August 2025, and Anthropic's help pages currently disagree about which models it covers. The usage API reports the scope per model, so Headroom names the row after whatever Anthropic is scoping that week instead of assuming it's Opus. If Anthropic stops reporting the window, the row fades to a dated ghost — “not reported since” and the date — rather than silently vanishing.
How do I check my Claude usage without opening claude.ai?
Anthropic's own readouts are Settings → Usage on claude.ai and the /usage command inside Claude Code, and they're accurate — they just live behind a tab and a terminal. Headroom reads the same provider-reported numbers and parks them in the Windows taskbar: percent left, reset countdown, and pace for the session window, the weekly window, and any model-scoped week your plan carries.
Claude isn't the only clock you're racing. Headroom also watches: