Claude Team pricing for a small business

Claude Team pricing for a small business

Claude Team pricing for a small business comes down to two seat prices and one minimum, and the minimum is lower than most people expect. As of July 2026 the Team plan starts at 2 seats, so a three-person shop qualifies with no sales conversation. A standard seat costs $20 per seat per month on annual billing, or $25 on monthly billing. A premium seat costs $100 per seat per month on annual billing, or $125 on monthly billing. Enterprise is closed to you at this size: the self-serve Enterprise tier starts at 20 seats and the sales-assisted tier starts at 50.

That narrows the decision to two questions. How many of your seats need to be premium? And which of your workloads should come off seats entirely and run on an API (application programming interface) key instead?

Every number below was on claude.com/pricing and in the Team plan documentation on 30 July 2026. Treat those pages as the source of truth, because this one will age. If you are still choosing between tiers rather than sizing one, start with which Claude plan you need.

What a Team seat buys, and what it does not

A standard seat gives about 1.25x the per-session usage of a Pro account. A premium seat gives about 6.25x. Both figures are measured against a single Pro account, so Claude Pro's own price and usage limits are the baseline worth knowing before you read either multiplier as generous. Both seat types include Claude on web, desktop and mobile, and both include Claude Code. If you are comparing that against paying for Claude Code on its own, what Claude Code costs outside a Team plan sets out the Pro, Max and API credit routes. If any of your people work on Linux, the desktop half of that list arrives by a different route, and what Claude runs natively on Linux covers the beta desktop package, the CLI and which paid plan covers each. Neither seat type buys a smarter model: every seat reaches the same model lineup, so you are paying for volume and for administration.

On the administration side, Team adds single sign-on (SSO) and domain capture, role-based access control, spending caps, and one central invoice instead of a pile of personal expense claims. It also settles the awkward question of whose credit card is paying, and it lets you cut off a leaver's access on their last day. Individual Pro accounts give you none of that. If the people joining are already paying for Pro themselves, cancelling or downgrading that personal plan keeps their access running to the end of the month they have already paid for, so nobody loses time they bought.

Now the part that catches small teams out. Usage limits on Team apply per member, and they are not pooled. Buying a seventh seat does not hand extra capacity to the developer who keeps running out. Nor does one person's allowance split by tool: Claude Code draws on the same usage window as the chat apps on every paid plan, so a developer's terminal sessions and their browser tab spend from one pot. Weekly limits reset on a fixed schedule for each person separately. If one person hits the wall every week, the fix is a premium seat for that person, or extra usage credits, not more headcount on the plan. In the meantime that person still has a day to salvage, and working out which limit window they have actually hit decides whether they wait out a reset, switch to a lighter model, or buy credits on the spot.

Worked example: a 5-person team

Take a five-person company: two developers who live in Claude Code all day, one designer, one marketer, and a founder who mostly chats.

Five standard seats on annual billing costs 5 x $20 = $100 per month. That is the floor, and it is the wrong shape. The two developers will hit their weekly cap by Thursday, because agentic coding sends the whole working context back to the model on every turn, and Claude Code's token usage is dominated by that re-sent context rather than by anything a person types.

Two premium seats for the developers plus three standard seats for everyone else, on annual billing, costs (2 x $100) + (3 x $20) = $260 per month. The same mix on monthly billing costs (2 x $125) + (3 x $25) = $325 per month, so committing annually saves $780 a year at this size.

Is $260 the right answer? Compare it to the individual route. Two Max accounts plus three Pro accounts costs more than $260 and gives you no admin surface at all. If you do price that route out, the choice between the $100 and $200 Max tiers rests on the same usage multiplier question a premium seat answers, just without the single invoice or the ability to cut off a leaver. That comparison assumes all five people need a paid account, which is worth testing one at a time: whether Claude Pro earns its keep for a given person turns on how often the free tier actually stops them, and a founder who dips in a few times a week may not clear that bar. If somebody is going to ask what those same five people would cost on ChatGPT instead, the per-person price comparison between the two settles it, including the coding case that matters most to your two developers. The premium seat is the cheap part of this decision. The expensive part is a developer stalled for two days a week because a standard seat ran dry. If you would rather see that argument as arithmetic than as assertion, the hours a Claude Code plan has to recover to pay for itself is a calculation you can run at your own developers' hourly rate.

Worked example: a 12-person team

Twelve people, four of them heavy Claude Code users, eight of them regular chat and document users.

Four premium seats plus eight standard seats on annual billing costs (4 x $100) + (8 x $20) = $560 per month, or $6,720 a year. The same mix on monthly billing costs $700 per month. The four premium seats are $400 of that bill, so whether metering the same terminal sessions per token would come in under $400 a month is worth putting next to the seat-versus-token comparison for Claude Code before you commit for a year.

At twelve people the Enterprise question comes up, and as of July 2026 the answer is no. Enterprise self-serve requires 20 seats. You would buy eight seats nobody uses just to qualify, which is 8 x $20 = $160 a month in seat fees alone, and then every token your team uses is billed separately at API rates on top of that. Enterprise makes sense when premium seat caps are the daily bottleneck for most of the team, or when a compliance requirement such as audit logs, SCIM provisioning or a HIPAA-ready configuration forces it. At twelve people it rarely is either. If you want to see how that bill assembles before you rule it out, what the Enterprise seat price actually covers walks through the seat minimums, the metered token line and the parts that only appear in a negotiated quote.

When an API key beats a seat

A seat is a flat monthly fee with a usage ceiling. The API is a meter with no ceiling. Nothing free sits underneath that meter either: there is no API free tier, only a small credit at signup, though a few parts of the platform genuinely cost nothing. The crossover is easier to compute than people assume.

A standard seat costs $20 a month on annual billing. At the Claude Sonnet 5 introductory API rate of $2 per million input tokens and $10 per million output tokens, in effect through 31 August 2026, that same $20 buys roughly 5 million input tokens plus 1 million output tokens. A genuinely occasional user, someone who opens Claude twice a week to rewrite an email, does not come close to that.

So on raw token cost the API wins for light users. Three things stop that being the obvious answer:

• An API key is not the Claude apps. It gives no chat interface, no projects, no shared knowledge bases, and no mobile access. A non-technical colleague cannot use a raw key, and building something for them to use costs more than the seat.
• API billing has no ceiling by default. Set a spend limit in the Console before you hand a key to anything that loops.
• Per-token spend moves month to month, which is exactly what a small business budget dislikes.

The working rule: buy seats for people, buy API access for programs. If a human at a keyboard starts the request, that is a seat. If a scheduled job, a webhook handler or an internal tool starts it, that is an API key. The full subscription versus API cost comparison works through the token math for the heavier cases.

The mixed pattern most small teams land on

The stable shape for a 3 to 15 person business is two billing relationships, not one.

Seats cover the humans: premium for the people in Claude Code daily, standard for everyone else, one invoice, SSO on top. An API key covers the machinery: the ticket classifier, the nightly report summariser, the internal search tool. Those run against a Console account, billed per token, under a spend cap, and they should use a cheaper model than your developers do. Most automation is classification and extraction, which Claude Haiku 4.5 handles at $1 per million input tokens and $5 per million output tokens as of July 2026, a fifth of the Opus rate. Picking the right model for each task saves more than shaving a seat off the plan. The exception is the job that is closer to drafting than to sorting, where the output is read by a customer rather than by another script, and Fable 5's published rates and the work that justifies them are worth checking before you assume the cheapest model on the list wins.

Two smaller levers matter once the tools are running. Batch processing halves both input and output rates for work that does not need an answer this second, such as an overnight summary run. Prompt caching charges a cache read at a tenth of the base input rate, which pays off as soon as the same system prompt or document is sent twice. Neither lever applies to seat billing, which is another reason to keep automation on the API side.

If those tools run on a server you control, the account setup is identical, and a first Claude API app on a VPS covers key handling and the first working script. Keep the key out of anything a person can read.

What to do this week

• Count the people who will use Claude every working day, not the people who might. Those are your premium seats.
• Put everyone else on standard seats. Upgrading somebody later takes a minute.
• Run one monthly cycle first to see the real usage shape, then switch to annual once the mix has settled. The annual discount is 20 percent on both seat types.
• Set a spending cap before you create an API key, not after the first surprise invoice.
• Check the prices yourself on the day you buy. Seat prices and API rates both changed in the last twelve months, and the Sonnet 5 introductory rate expires on 31 August 2026.

FAQ

What is the minimum number of seats on the Claude Team plan?

Two, as of July 2026, with a maximum of 150. A three-person or five-person business can buy Team directly, with no sales process. Seats can be mixed, so you are not forced onto the expensive seat type for people who only chat. Above 150 seats the plan stops scaling and Enterprise is the next step.

Can a 12-person company buy Claude Enterprise?

Not as of July 2026. Enterprise self-serve carries a 20-seat minimum and the sales-assisted route carries a 50-seat minimum, so a twelve-person company would have to buy eight seats nobody uses. Enterprise also bills on a different model: the seat fee, around $20 per seat per month billed annually, covers access only, and every token the team uses is billed separately at API rates with no per-seat ceiling.

Is a Claude Team seat cheaper than paying for the API?

For a daily user, yes, by a wide margin. For an occasional user, no. A $20 standard seat costs the same as roughly 5 million input plus 1 million output tokens at Claude Sonnet 5 introductory API rates as of July 2026, and a light user rarely reaches that volume. The catch is that an API key provides no chat apps, no projects and no mobile access, so the saving is only real when the person is technical or the usage sits inside a tool you built.

Does buying more Team seats give my heaviest user more usage?

No. Team usage limits apply per member and are not pooled, so an extra seat is one more person's allowance rather than a shared pot. If one developer keeps hitting the weekly cap, move that person to a premium seat, which carries about 6.25x Pro's per-session usage against 1.25x on a standard seat, or buy additional usage credits for that account.

Should a small team pick annual or monthly billing for Claude Team?

Run one monthly cycle first. Monthly billing costs $25 per standard seat and $125 per premium seat, against $20 and $100 on annual billing, so annual saves 20 percent. That discount is worth taking only once you know how many premium seats you really need. Locking a guessed seat mix in for a year is the more expensive mistake.