AI seats for teams

How to read a seat row #

A seat price is three numbers, not one. (1) The rate per seat — and which billing term earns it: Claude Team's $20 is the annual-billed rate, $25 paid monthly (Anthropic Claude Team - Standard seat row, DIRECT, accessed 2026-09-07). (2) The range the seller states: the same row reads 2–150 seats. (3) Anything fixed on top: Devin's Teams entry is $80 plus $40/seat (Windsurf / Devin (Cognition) Devin pricing (post-merge ladder) row, EXCERPT, accessed 2026-09-07). Multiply the rate by headcount and you have the line item; the other two numbers decide whether that line item survives contact with your org chart.

Worked seat math #

Mixed Claude Team seats #

The row for Premium seats states 5x Standard-seat usage and mixing with Standard seats (Anthropic Claude Team - Premium seat row, DIRECT, accessed 2026-09-07). Two heavy users on Standard seats and one on Premium, at annual rates:

Same team, all Standard: 3 × $20 = $60/mo annual — $80/mo cheaper than the mix. The mix is worth it only if one person's usage genuinely needs the 5x tier; otherwise you are renting headroom.

Three-person office, one seat class #

Three Standard Claude Team seats: $60/mo on the annual rate, $75/mo billed monthly (Anthropic Claude Team - Standard seat row, DIRECT, accessed 2026-09-07). For reference, one consumer Claude Pro seat runs $20/mo monthly (Claude Pro row, accessed 2026-09-07) — three Pro accounts match the annual-billed rate on three Team seats, but give up the team workspace and the mix-and-match seat classes, and the Team row states a 2-seat minimum.

Copilot: five seats, five credit buckets #

5 × $19 = $95/mo (GitHub Copilot Business row, DIRECT, accessed 2026-09-07). The included 1,900 AI credits per user per month is stated per user per month — the row never describes a team pool — so a heavy developer burns their bucket while teammates' sit idle, and extra credits bill at $0.01 per AI credit per the row's terms. Copilot Enterprise doubles the seat fee and the bucket ($39/seat, 3,900 AI credits per user per month) (GitHub Copilot Enterprise row, DIRECT, accessed 2026-09-07).

Devin: the base fee is the small-team trap #

Teams is $80/mo + $40 per seat (Windsurf / Devin (Cognition) Devin pricing (post-merge ladder) row, EXCERPT, accessed 2026-09-07): 5 seats = $280/mo, but at two seats the base fee alone makes it $80 per seat — the fixed $80 is half the bill. The row's own caveat says its pricing came from a page snapshot.

Cursor: seat class is the usage dial #

5 Teams seats: $200/mo on Standard or $600/mo on Premium, which the row defines as 5x Standard Agent limits (Cursor Teams Standard / Premium row, DIRECT, accessed 2026-09-07). Upgrading a seat, not adding one, is how this row scales heavy users — and the row notes an Enterprise tier “custom with pooled usage,” i.e. past teams, the model switches to pool pricing.

Decision rules: seats vs pooled API spend #

1. Named humans in apps → seats. Seats exist to give people a product (chat UI, Claude Code, Copilot in the IDE). A shared API key for a chat-tool team trades away the seat features you were buying — and usually violates the consumer side of the terms you're reading.
2. Programs calling models → pooled spend. Bots, batch jobs, and internal tool backends don't need a seat; buy metered tokens and let one bill absorb everyone. The break-even panel in the calculators prices task volume against a flat rate.
3. Uneven usage → mix seat classes, don't max them. The $140/mo example above is the pattern: the row explicitly allows it on Claude Team. Buying everyone the tier their loudest week needs is the classic team overspend.
4. Know which wall you're buying. Both Claude Team seat rows state no per-seat overage figure (checked 2026-09-07), and the Standard row is defined against the consumer tier's caps — “more usage than pro” while the Pro row carries 5-hour session caps. On those rows the heavy month is a usage-limit question, not an invoice. On rows that publish metered excess — Claude Enterprise (“$20/seat/mo plus usage billed at API rates” (Anthropic Claude Enterprise row, DIRECT, accessed 2026-09-07)), Copilot credits at $0.01 per AI credit, Devin's “extra usage bills at API pricing,” Mistral's Vibe Team PAYG credits — a hungry teammate is a bill. Budget seats-only or seats-plus-usage-cap, deliberately.
5. Compare per head, not per plan. A $20seat used daily for real work beats a $0 API wallet nobody remembers to fund; a seat nobody logs into is pure loss. If more than a third of your seats would sit idle, price a partial-seat setup (heavier staff seated, rest on pooled API or free tiers) before committing to annual billing.
6. Annual seat rates are a headcount bet. The $20 vs $25 Claude delta is real only while the roster is stable — the seat calculator totals both terms side by side; read the lock-in checklist in monthly vs annual AI plans.

Where the data stops #

• ChatGPT Business: no public per-seat dollar. The verified structure is per-seat limits plus a shared credit pool, and the row also notes Codex seats closed to new Business workspaces from Jun 24 2026 (OpenAI ChatGPT Business seat row, UNCERTAIN, accessed 2026-09-07) — structure yes, arithmetic no.
• Claude Team seat overage: not published. Both seat rows were checked against their source on 2026-09-07; no per-unit excess figure exists on the row, so heavy-seat months are a wall question, not a bill one, per current terms.
• Google seats aren't in the full price table. The closest Google row is the consumer AI Pro plan at $19.99/mo (accessed 2026-09-07); we don't restate a consumer price as a seat price, so no Google entry appears in the seat calculator.

Every tracked seat SKU, with the full caveats, sits in the For teams table on the offers page; pick a SKU and a headcount in the seat calculator.

Prices used in this guide #

Row sources are the provider pages linked from each provider's page; full price table snapshot 2026-09-07.

FAQ #