Updated guide · June 2026
Cowork does not get its own allowance. It is an agentic surface that draws from the very same usage pool as Claude.ai chat and Claude Code, so a heavy Cowork session quietly spends the budget you might have wanted for chatting later. This guide explains how Cowork interacts with the five-hour and weekly limits, what burns usage fast, and how to stretch a shared plan.
Straight answer
Cowork has no separate quota — it draws from your plan's shared 5-hour + weekly usage with Claude chat and Claude Code. There is no extra Cowork allowance to buy or track; everything you do in Cowork counts against the same pool as everything else on your subscription. Verify in Claude → Settings → Usage; numbers change.
Chat, Claude Code and Cowork all spend the same five-hour and weekly budget. Because Cowork is agentic and does a lot of work per task, a single Cowork run can use much more of that pool than a handful of chat messages. Verify in Claude → Settings → Usage; numbers change.
The shared pool
The mental model that makes Cowork limits make sense is simple: your subscription gives you one budget, and chat, Claude Code and Cowork all spend from it. There is no separate Cowork meter. Whatever you do in one surface leaves less for the others in the same window.
You cannot buy or earn a dedicated Cowork quota. Cowork usage is simply part of your plan's overall five-hour and weekly budget, the same one chat and Claude Code draw from.
Cowork reads files, runs tool calls, reasons across multiple steps and iterates toward a goal. One Cowork task can therefore consume far more of the pool than a short chat exchange.
If a long Cowork run earlier in a window uses most of your budget, Claude may slow or pause you in chat afterward even though you barely typed anything. The pool is shared, so the order of work matters.
Because everything draws from one pool, Settings then Usage reflects the combined effect of chat, Claude Code and Cowork. That screen, not any fixed number, tells you where you stand.
The two layers
Cowork is subject to the same two stacked limits as the rest of your plan. Knowing which layer a Cowork session is pressing on tells you whether a short wait clears it or whether you need to rethink your week.
| Layer | How Cowork affects it | How it resets | What to do |
|---|---|---|---|
| Five-hour session | A long agentic Cowork run can spend most of the rolling window quickly because it does a lot of work per task. | Resets about five hours after your first message of the window, then a new window begins. | If only this is hit, a short wait usually clears it. Sequence big Cowork runs into focused bursts. |
| Weekly limit | Repeated heavy Cowork sessions across the week can use up the longer weekly budget shared with chat and Code. | On its own weekly schedule shown in Settings then Usage, with separate Opus versus other-models reset. | If this is hit, waiting hours will not help. Lighten model use, narrow scope, or consider a higher plan. |
To see exactly where a Cowork session left you, open Claude, go to Settings then Usage, and read both the session and weekly indicators. Treat that screen as the source of truth, not this page.
Burn rate
Because Cowork is agentic, the gap between a cheap task and an expensive one is large. These habits keep a Cowork session from quietly draining the budget you share with chat and Claude Code.
Give Cowork a clear, bounded goal. A vague task makes it explore, read extra files and iterate more, all of which spend more of the shared pool.
Where the task allows, prefer a lighter model. Heavier models do more work per step and draw more from your budget, just as they do in chat.
Re-attaching the same large files on every step reloads them each time. Point Cowork at material once rather than pasting big files repeatedly.
Long, sprawling agent sessions carry more context and iterate more. Break big jobs into smaller, well-defined runs so each one stays cheap.
Run heavy Cowork tasks when you do not also need a lot of chat headroom in the same window, since they share one pool.
Check Settings then Usage before and after big Cowork runs so you understand how much of the shared budget each one cost.
A day with Cowork
Because Cowork has no published quota, the clearest way to understand it is to compare two realistic days on the same plan. The difference is not the plan; it is how the shared pool is spent.
You hand Cowork a couple of well-scoped tasks: tidy a folder, draft a short document from notes, summarize one report. Each run is bounded, you point it at files once, and you use a lighter model where you can. In between you chat normally. Over the day Cowork never dominates the pool, the five-hour window resets quietly between bursts, and your weekly budget barely moves. This is the comfortable way Cowork is meant to feel.
You launch a sprawling Cowork session that reads dozens of files, iterates many times on a complex build, and runs for an extended stretch on a heavy model. Then you switch to chat expecting your usual headroom, but Claude slows or pauses you. The culprit is not a Cowork-specific cap; it is that the one big agentic run spent most of the shared five-hour budget that chat and Claude Code also rely on.
Nothing in the heavy day is wasted, but it spends the shared pool many times faster than the light day. If your real work looks like the heavy day most days, that is the signal to tighten scope or move to a higher plan, not to assume Cowork is broken.
What counts
Cowork usage, like the rest of your plan, is measured in work rather than messages. The more files, reasoning steps and tool calls a task involves, the more of the shared pool it spends. This table shows the main contributors.
| Activity | Relative cost | Why |
|---|---|---|
| Long agentic runs | High | Many reasoning steps and iterations add up far beyond a single chat reply. |
| Reading many files | High | Each file adds context that is processed, and large or numerous files compound it. |
| Heavy models | High | The most capable models do the most work per step and draw more from the pool. |
| Tool and web calls | Medium to high | Fetching, searching and running tools adds processing on top of the reasoning. |
| Re-uploading the same files | Medium to high | Reloading the same material on every step repeats the context cost each time. |
| Lighter, scoped tasks | Low to medium | A bounded task on a lighter model finishes in fewer steps and spends less. |
None of these are off-limits in Cowork. The point is awareness: when you understand that a long multi-file agent run is dozens of times heavier than a short chat, the pauses in chat afterward stop feeling random.
Symptoms
Most surprise pauses around Cowork trace back to the shared pool. Match what you saw before assuming something is wrong with your account.
A big agentic session earlier spent most of the shared five-hour budget, leaving little for chat. This is expected with one pool. A short wait for the window to reset usually restores chat.
A long, sprawling run carries growing context and iterates repeatedly, so it costs more as it goes. Breaking the job into smaller, scoped runs keeps each one cheap.
If a short wait does not help, you are on the weekly limit, shared across chat, Code and Cowork. It resets on its own schedule in Settings then Usage, so the fix is lighter use or a higher plan, not waiting.
If you barely chatted but still hit a wall, agentic work in Cowork or Claude Code is the usual hidden cause, because all three draw from the same budget. Check Settings then Usage to confirm.
Alternatives
If a heavy Cowork run leaves you capped for chat, the cleanest fix is not depending on one app's pool. MultipleChat lets you run Claude alongside other models, so a single provider's limit does not block your work.
Run Claude-style answers next to ChatGPT, Gemini and search models in one workspace. When your Claude pool is spent by a Cowork run, keep working in another model instead of waiting.
AlternativeBroad assistant for writing, coding, images and everyday productivity when Claude is unavailable.
AlternativeUseful for Google Workspace and Google-context work.
ResearchResearch-focused answers with web sources when you need citations.
Our guideThe other agentic surface that shares the same pool.
Our guideHow the five-hour and weekly limits work on Pro.
Our guideThe separate per-minute limits and tiers on the API.
Sources
Limits, prices and behavior change. Use these official pages and your own Settings then Usage screen before making decisions.
Open Claude and check Settings then Usage for your live session and weekly limits.
Official plan names, current pricing and usage wording.
Details on what consumes usage across surfaces.
How agentic work shares your plan usage pool.
The usage policy that applies across all Claude surfaces.
FAQ
Short, hedged answers for people searching Claude Cowork limit, Cowork usage, shared pool and Cowork vs Claude Code.
No. Cowork has no separate quota. It draws from the same plan usage pool as Claude.ai chat and Claude Code, governed by your rolling five-hour limit plus weekly limits. Heavy Cowork work reduces what is left for chat and coding in the same window. Verify your live usage in Claude under Settings then Usage.
Cowork activity counts against the same rolling five-hour session limit and the weekly limits as the rest of your plan. A long Cowork run can pause you for chat, and a heavy Cowork week can use up your weekly budget. Both reset on their normal schedules shown in Settings then Usage.
Cowork is agentic. It reads and writes files, runs multiple reasoning steps, calls tools and iterates on a task, so a single Cowork session can do far more work than a short chat reply. More work per task means more usage drawn from your shared pool. Watch Settings then Usage to see the effect.
Yes. Because Cowork shares one usage pool with chat and Claude Code, a big Cowork run earlier in a window can leave little headroom for chat afterward, so Claude may slow or pause you even though you barely chatted. Check Settings then Usage to confirm which layer paused you.
Cowork is part of the Claude subscription experience and uses your plan's shared usage rather than a separate allowance. The amount of Cowork work you can do scales with your plan tier, since Pro, Max, Team and Enterprise have different five-hour and weekly headroom. Verify availability and limits on the official Claude pages.
Scope tasks tightly, give clear instructions so Cowork does not wander, avoid re-uploading the same large files, prefer lighter models where the task allows, and sequence big Cowork runs for times you do not also need a lot of chat headroom. These habits stretch the shared pool.
Both Cowork and Claude Code are agentic surfaces that draw from the same shared subscription pool as chat. They are different interfaces, but neither has a separate quota on a subscription, so heavy use of either reduces what is left for the others in the same window. Verify behavior on the official support article.
Open Claude, go to Settings then Usage, and read your session and weekly indicators. Because Cowork shares that pool, the same screen shows the combined effect of chat, Claude Code and Cowork. Treat that screen as the source of truth rather than any fixed number.