One number is circulating for Claude Cowork, and it is $100. It is not wrong, which is what makes it useless: it names one rung of a ladder that starts at $20. So here is the honest version: Cowork has no subscription of its own, your existing Claude plan already includes it, and the thing that decides your cost is not the tier label. It is how fast your multi-step tasks eat through a shared usage pool that nobody publishes as a number. Which plan are you on right now, and how much of it is your agentic work quietly consuming every week?
The limit that decides your Claude Cowork price is not printed anywhere as a count. Not a task number, not a credit number. Anthropic tells you what moves it instead of what it is, and that gap is where a plan choice gets made on the wrong signal. You can pick one, run it for three weeks, and discover that a single afternoon of file-heavy automation costs more of your allowance than a month of chat. Or you can go the other way and pay for headroom you never touch. Both mistakes are expensive and only one of them is obvious. The correct choice depends on the shape of your work, and the vendor's own product page states the mechanism in a single sentence that turns out to be the whole argument.
Which Claude Cowork plan you need before the limit bites
No vendor surface publishes a task count for Cowork, so the field invents one or ignores the question. Both are worse than arithmetic. What Anthropic does publish is a set of levers that convert directly into a plan recommendation, and they are enough to work with.
Three facts do most of the work:
One pool, every surface. Your usage across Claude on web, desktop, mobile, Claude Code and Cowork all draws from the same allowance. Cowork is not a separate meter you can spend freely while chat sits idle.
Two clocks. A rolling five-hour session window resets through the day, and paid plans add weekly limits on top. The weekly cap is the one that ruins a Monday, because it does not care that you had a quiet weekend.
Steps cost more than sentences. Anthropic's own help documentation states that multi-step tasks which run code, create files, or use your connected apps and browser use more usage than a quick question, because each step Claude takes consumes tokens. Cowork's product page puts it bluntly: Cowork consumes limits faster than chat.
Read those three together and the mapping writes itself. What matters is steps per task, multiplied by how many surfaces you are running at once. Not tasks per month.
Your working pattern | What it does to the shared pool | Plan that holds |
|---|---|---|
Occasional chat, one folder tidy-up a week, short reports | Light draw, one surface active at a time | Pro |
Daily asset work: a few files produced per day, some browsing, chat running alongside | Steady draw, session window is the binding constraint rather than the week | Pro for most weeks, with overage in heavy ones |
Handing off work throughout the day: long multi-step jobs, code or file creation, connected apps in the loop | Heavy draw. Session and weekly caps both bite, and chat competes with Cowork for the same allowance | Max 5x |
Parallel jobs running most of the working day, several surfaces at once | Sustained ceiling pressure. Weekly caps arrive mid-week and stay | Max 20x |
A team where only one or two people do the heavy agentic work | One seat shoulders the load, the rest are light | Mixed Team seats: standard for the light users, premium for the heavy one |
The pattern column is a description, not a measurement. Anthropic publishes no per-task figure and we are not going to invent one. What is measured here is the mechanism: shared pool, two clocks, per-step consumption, and relative multipliers that are stated against different baselines. Max 5x and Max 20x give five and twenty times the Pro allowance per session. Pro itself gives at least five times the free service per session. Note that asymmetry, because it means Pro is the reference point for every upgrade decision you will make.
Claude Cowork pricing, plan by plan, in dollars
Every figure below was read from claude.com/pricing and Anthropic's help centre on October 1, 2026, in USD. Prices exclude tax and Anthropic states that price and plans can change at its discretion. If you are reading this months later, re-check the vendor page before you budget.
Plan | Billed monthly | Billed annually | What a year costs you | Annual option exists? |
|---|---|---|---|---|
Free | $0 | $0 | $0 | Not applicable |
Pro | $20 per month | $200 billed up front, which the vendor renders as $17 per month effective | $200 | Yes |
Max 5x | $100 per month | Not offered | $1,200 (12 x monthly) | No |
Max 20x | $200 per month | Not offered | $2,400 (12 x monthly) | No |
Team, standard seat | $25 per seat per month | $20 per seat per month | $240 per seat | Yes |
Team, premium seat | $125 per seat per month | $100 per seat per month | $1,200 per seat | Yes |
Enterprise | Not offered monthly | $20 per seat per month plus usage at API rates | $240 per seat plus usage | Annual only |
Two rows deserve their own paragraph, because both break assumptions people bring from other tools.
Max has no annual billing at all. Anthropic's FAQ is explicit that annual subscriptions exist for Pro and Team, and that Enterprise is available only as an annual subscription. Max is monthly in both bands. So there is no discounted Max yearly figure to find, and any page showing you one has done the multiplication and presented it as a plan. The honest way to state a Max budget is $1,200 a year for 5x and $2,400 for 20x, labelled as arithmetic.
Team is per seat, and the two seat types are five times apart. A standard seat costs $20 a year-billed, a premium seat $100, and Anthropic allows the two to be mixed in the same organization for teams of 2 to 150 people. If your load sits with one person, that mixing is where the savings are, and we come back to it in the setups below. If you are comparing this shape to another vendor's per-seat model, another seat-based plan ladder runs through the same arithmetic, and how another assistant prices its tiers shows a simpler version of it.
The Cowork usage limit, explained honestly
Here is what the vendor does document, and where the documentation stops.
Your five-hour session limit and your weekly limits are visible in Settings, under Usage, as progress bars with the next reset time. That is the only place your real number exists, and it is per account. Weeklies reset on a fixed day and time for your account, and on Max they apply across all models.
What consumes the pool is a longer list than you probably expect: the length of your messages, the size of attached files, how long the current conversation has run, tool use such as research and web search, which model you picked, the effort level, artifact creation, and multi-step work such as running code, creating files, or browsing. Cowork's whole value proposition sits in the last item on that list, which is exactly why it draws down faster than chatting.
Anthropic publishes four mitigations, all of which are free:
- Group related work into a single task instead of firing several separate ones.
- Start a new conversation for unrelated work, so Claude is not carrying context you no longer need.
- Say when you want a quick answer rather than a finished deliverable, and the agent stops short of building the artifact.
- Keep reusable material in Projects, because cached content does not count against limits when it is reused.
If you are already on the top tier, or you cannot move tiers mid-quarter, there is a second mechanism almost nobody mentions in this context, and it is the difference between a hard stop and a soft one.
When the next tier up costs less than working around the limit
Once your plan limit is reached, you have three options rather than two. Wait for the reset, move to a higher plan, or turn on usage credits and keep working at standard API rates. The third option is available on Pro, Max 5x, Max 20x and Team, and it changes the arithmetic that decides whether an upgrade is worth it.
Anthropic sells pre-purchased usage bundles with a published discount ladder, and the same balance works across Claude, Claude Desktop, Claude Mobile, Claude Code, Cowork and third-party products on your account:
Bundle | Discount | You pay |
|---|---|---|
$50 | 10% | $45 |
$250 | 20% | $200 |
$1000 | 30% | $700 |
Individual Pro and Max subscribers can buy up to $2,000 of discounted bundles per month, and Team owners up to $3,000, after which usage bills at standard rates. The bundle credits only apply after your plan's included usage is exhausted, so they extend your ceiling rather than replacing it.
That pricing makes the crossover computable, so here it is per tier, labelled as our arithmetic on Anthropic's published numbers rather than a vendor figure:
Pro to Max 5x. Moving up costs $1,000 more a year ($1,200 against $200). If the usage you would otherwise buy beyond Pro's allowance costs you more than $1,000 a year at the best bundle discount, the upgrade is the cheaper route. That threshold is about $1,429 of usage at list price, which the 30% discount brings down to $1,000. Below it, stay on Pro and buy the occasional bundle.
Max 5x to Max 20x. The gap is $1,200 a year, so the same test applies with a higher bar: once discounted bundle spend on 5x passes $1,200, which is roughly $1,714 of usage at list, 20x costs less than the top-ups it replaces.
Team standard to Team premium. The gap is $960 a year per seat ($1,200 against $240), and the premium seat gives five times the standard seat's usage. If one person on the team is drawing more than about $960 a year of discounted overage on a standard seat, that seat should move up, and about $1,371 of list-price usage is where the crossover sits. This is the reason to mix seat types rather than standardize.
Below Pro, there is no crossover. The free plan does not include Cowork, so there is nothing to straddle.
Three setups, priced
Costs below use annual billing wherever the tier offers it, because that is the cheaper way to hold a tier you intend to keep.
Solo marketer running daily asset work. A few deliverables a day, chat open most of the morning, occasional browsing and file creation. This fits Pro with room to spare in a normal week, and a $250 bundle or two covers the months when a launch lands. Annual figure: $200, plus overage as needed. For context on what the tool actually does across a working week, our Claude AI review covers that ground, and AI writing work inside one workspace is the version of this job that never touches a usage meter.
Three-person content team. One writer producing most of the output, two people using Claude for review and planning. Standard Team seats for all three is the correct shape, with the heavy user's overage handled by bundles. Annual figure: $720 for three standard seats, plus overage.
Agency running client work in parallel. One operator running long multi-step jobs across multiple clients all day, two or three lighter users supporting. Standardizing on premium seats wastes money here, and standardizing on standard seats stalls the operator. Mix them: one premium seat plus three standard seats is $1,200 + $720 = $1,920 a year, and it is the configuration that matches where the load actually sits. If the agency also has engineering-led work in flight, the developer-facing plan question is worth reading before you commit those seats, and which Claude surface to use settles the adjacent decision.
Is Claude Cowork free?
Today, no. As of October 2026, Cowork is not on the free plan. Anthropic's plan page does not list it there, and its rollout language puts free access in the "coming" bucket rather than the current one.
That answer has a shelf life, and it is short. On September 16, 2026, Anthropic merged Claude Cowork and chat into a single Claude, and the announcement states the rollout is going to Pro and Max first, with Team and Free plans to follow. The help centre describes the same staging: some accounts still show separate Chat and Cowork options while the change lands, and once your account has the merged experience you cannot switch back. So the accurate sentence is that Cowork is a paid-plan capability right now, with free access announced but not delivered. Anyone stating it as a permanent exclusion is writing about a product that moved six weeks ago.
The bottom line
One recommendation per setup. Solo marketer on daily asset work: Pro, $200 a year, plus a bundle when a launch lands. Three-person content team: three standard Team seats, $720 a year, and move the heaviest seat to premium only if its overage passes roughly $960 a year. Agency running parallel client work: one premium seat and three standard seats, $1,920 a year, because a tier-wide upgrade pays for headroom most of the team will not use.
The framing worth keeping is the one the field skips. Agentic desktop work is one input into a pipeline that also has to produce, publish and measure something, and a Claude plan buys you that input and nothing further. If the plan cost is the whole budget, the arithmetic is clean. If it is one line of a stack that also carries a keyword tool, a scheduler and a way to get the work live, then the honest comparison is against what the whole stack produces, not against a tier label. That comparison is the one your CFO will actually make.
Frequently asked questions
How much is Claude Cowork in total, and what does that cost cover?
Nothing on top of the plan you already hold. Claude Cowork cost is not a line item, because Cowork has no separate subscription and no add-on price. Your cost is the tier you need in order to keep working, which is $200 a year on Pro, $1,200 on Max 5x or $240 per standard Team seat. If you exceed your plan's allowance, the additional cost is usage credits at standard API rates, or pre-purchased bundles at up to 30% off.
Which Claude plan includes Cowork?
Pro, Max 5x, Max 20x, Team on both standard and premium seats, and Enterprise. Those are the Claude Cowork plans that ship it today: the tier you pick is a seats and usage decision, not a Cowork purchase. The free plan is the only tier where Cowork is absent today, and Anthropic has said Team and Free will follow.
Is the Claude Cowork Pro plan enough for me?
It is enough for light and moderate patterns: a few deliverables a day, single tasks rather than parallel ones. It stops being enough the moment you are handing off long multi-step jobs repeatedly, because Pro's session allowance is the baseline that Max multiplies rather than a large pool in its own right. Check Settings, then Usage, and watch whether your five-hour bar is filling before your week does.
Does Cowork need its own subscription during the rollout?
No. The September 16, 2026 merge folds Cowork into the same Claude you already use, and Cowork's own product page presents it as included in each plan rather than sold separately. There is no Claude Cowork Pro subscription to buy in addition to your Claude Pro one. You keep your existing plan, seats and billing. Nothing is repurchased.
What happens if I hit the Cowork usage limit?
You have three options. Wait for the five-hour window or the weekly cap to reset, move to a higher plan, or enable usage credits and continue at standard API rates. Bundles bought in advance reduce the third option by 10% to 30% depending on size, and the balance is shared across every Claude surface and third-party product on your account.
Is annual billing cheaper for Cowork?
Yes on Pro, Team and Enterprise, and not available on Max. Pro at $200 up front saves $40 against twelve monthly payments. Team annual billing drops a standard seat from $25 to $20 per month and a premium seat from $125 to $100. Max is monthly in both bands, so a yearly Max figure is multiplication rather than a discount.
Does Cowork share a limit with Claude Code?
Yes. Work in the terminal, on the web, in the desktop app, on mobile and in Cowork all draw from one pool, and Anthropic's own FAQ names Claude Code as sharing the same limits as the rest of your plan. If several people or surfaces are active at once, they are competing for the same allowance rather than adding to it.
Are prices the same outside the United States?
No. Anthropic states that pricing varies by region and that local currency applies where supported. Every figure in this article is USD and was verified against the US pricing surface on October 1, 2026.
Start with the pipeline, not just the plan
Allable runs the whole chain in one workspace, from keyword research through writing, design and publishing, so the output of an agentic session lands somewhere instead of in a clipboard. Start with a 7-day free trial and cancel any time.
