First, the names

People ask for "Claude Business" and there is no such thing. Anthropic's team-facing subscription is Claude Team, with Claude Enterprise above it, both used through the Claude app. The developer product is the Claude Platform at platform.claude.com - the API, the SDKs, and the Claude Console, the web surface where its organization, workspaces, keys, spend limits, and billing live. Same company, same sign-in email, two separate accounts with separate billing.

This article is about the accounts, so when it says "the Console" it means that account layer - the thing you compare against a Team subscription. What the Platform does once you have a key (the Messages API, the agent loop, managed agents) is a different question, answered in What is the Claude Platform?.

Claude Team is a subscription for people. The Console is an account for programs.- the distinction to start from

If you have read ChatGPT Business vs OpenAI Platform, the shape is familiar: a per-seat app product on one side, a per-token developer platform on the other, and a subscription that buys nothing on the far side of the line. What is different here is that Anthropic puts doors in that line on purpose - which is either convenient or confusing depending on whether you knew about them.

CLAUDE.COM

Claude Team

The team tier of the Claude app: shared administration, seat management, org-level settings, and business data terms. Claude Enterprise sits above it with SSO, SCIM, audit logs, and compliance controls.

Bought for
Employees using Claude directly - chat, documents, analysis, and Claude Code in the terminal.
PLATFORM.CLAUDE.COM

Claude Platform

The developer product. Its Console holds an organization containing workspaces, each with its own API keys, spend limits, and usage reporting, plus the Workbench and the Admin API.

Bought for
Software calling Claude - your application, your agent, your pipeline, your CI job.

Claude Team

Claude Team is a seat-based subscription to the Claude app. You are buying access for named people plus the administration around them, not capacity your code can address.

What it gives you:

  • Seats, in more than one size. Team offers standard and premium seat types with different usage allowances, and you can mix them in one organization - not everyone needs the heavier tier.
  • Central administration. One organization, one bill, seat management, and org-level settings instead of a pile of individual subscriptions.
  • Claude Code on every seat. Anthropic ships Claude Code inside the paid plans, so the terminal tool is part of the subscription rather than something you reach only through the API.
  • Business data terms. The usual reason companies consolidate off personal accounts.
  • An Enterprise tier above it for SSO and domain capture, role-based access, SCIM provisioning, audit logs, and a compliance API - bought when governance, not capability, is the constraint.

Seat prices, seat minimums, and the per-plan usage multiples have all moved more than once, so treat any number you read - here or anywhere - as a snapshot and check Anthropic's pricing page before you budget. What has been stable is the shape: fixed cost per person per month, with usage limits attached to the seat rather than metered to a bill.

What Team does not give you is an API key. There is nothing in a Team organization to paste into a .env file.

Claude Platform

The Platform is the developer product; the Console is its account surface, and the account is what this section is about. You create an organization, and inside it you get workspaces - collaborative spaces that let you separate API resources by use case, team, or environment.

What matters about it in practice:

  • Workspaces scope everything. Keys, members, spend limits, and usage reporting all hang off a workspace, so staging and production can be genuinely separate. API keys are tied to the workspace they were created in and cannot be moved between workspaces.
  • Granular organization roles. Not just admin and member - the role list distinguishes who can use the Workbench, who can manage keys, who can touch billing, and who can manage users.
  • Usage-based billing. Per token, by model, from prepaid credits or a card on file. No seats, so no per-head component at all.
  • Rate limits by tier. Requests and tokens per minute depend on your account tier, visible in the Console, with per-model-tier limits settable per workspace.
  • Service tiers for throughput. Standard for everyday use, Priority when latency and availability matter, and Batch for asynchronous work at a discount.
  • An Admin API. Members, invites, workspaces, and keys are all manageable programmatically with a dedicated admin key, which is how onboarding and offboarding get automated rather than clicked.

The cost model is the thing to internalise. A seat is flat and predictable; token spend is a function of traffic, so the Console bill moves with what your software does. Anthropic has been publishing per-developer figures for Claude Code deployments precisely because that variance is hard to budget - the same problem covered in what an agent costs.

Side by side

Claude Team Claude Platform
Website claude.com platform.claude.com
Main purpose Employees use Claude directly Software calls Claude models
Interface The Claude app, plus Claude Code API, Workbench, dashboard, Admin API
Grouping Organization of seats Organization of workspaces
People Members on standard or premium seats Organization members with roles, added to workspaces
Billing Per seat, per month Per token, by model, from credits or a card
Spend control Seat count, and which seat type each member gets Per-workspace limits, rate limits, service tier choice
API keys No Yes, scoped to a workspace
Claude Code Included on paid seats Also usable, billed as API consumption
Tier above it Claude Enterprise - SSO, SCIM, audit logs, compliance API Higher rate-limit tiers, Priority capacity, custom terms

The row worth pausing on is Claude Code. It is the one capability that legitimately appears on both sides of the table, reached two different ways and billed two different ways - and that is the single most common source of "which account is this coming out of?"

Where the boundary leaks

The default rule is the same as OpenAI's: the two accounts are separate. But Anthropic has added deliberate crossings, and knowing which is which is the whole point of this section.

DOES NOT TRANSFER
  • The subscription. A paid Claude plan enhances the app experience; it does not include API or Console access. You set the Console up separately.
  • Membership. Team members are not Console organization members. Two invitations, two access lists, two sets of roles.
  • Keys. No key exists on the subscription side to find, and Console keys cannot even move between workspaces, let alone into the app.
  • Invoices. Separate billing, separate receipts, separate owners for finance to chase.
  • Rate limits. Console tier limits have nothing to do with your Claude plan's usage allowance.
DOES CROSS, BY DESIGN
  • Claude Code sits inside the subscription. It is included on paid plans and draws on the same plan limits as everything else, so terminal work and chat come out of one pool.
  • API rates reach into the plan. When you exhaust a paid plan's limits you can enable usage credits and keep working at standard API rates - metering, on the subscription side.
  • Heavy sessions can switch billing. For sustained coding work you can move to pay-as-you-go API credits through a Console account instead of waiting for a plan limit to reset.
  • Enterprise blends the two models. The Enterprise tier pairs a per-seat fee with usage billed at API rates, so cost scales with the models and tasks a team actually runs.

So the honest summary: OpenAI's two products are two walls; Anthropic's are two doors. The seat-versus-token distinction is identical, but on the Claude side token-rate billing can legitimately appear on the subscription invoice, and a developer tool ships with the seat. That flexibility is useful and it makes the question "which account paid for this?" a real one that someone has to own.

The word "workspace" is a trap

If your team uses both vendors, this is the detail that causes the most cross-talk in meetings. The same word means different things:

ChatGPT workspace
The team account itself - the container for members, seats, and the subscription. One per company, typically.
Claude Console workspace
A subdivision inside the developer organization - closer to an OpenAI project than to a ChatGPT workspace. You have many: one per app, team, or environment.
Claude Team organization
The subscription-side container for members and seats - the actual counterpart to a ChatGPT workspace.

Practical translation for anyone moving between the two platforms: a Claude Console workspace is an OpenAI project, and a Claude Team organization is a ChatGPT workspace. Getting those two pairs straight removes most of the confusion in a single step.

Which one you need

Same test as always - is the thing doing the asking a person or a program?

  • People need to use Claude at work. Claude Team. Seats, central admin, business terms, and Claude Code included for the developers among them.
  • Software needs to call Claude. The Claude Platform. A Console organization, a workspace per environment, keys scoped to each, and spend limits set before you ship.
  • Governance is the constraint, not capability. Claude Enterprise - and note it prices as seats plus usage at API rates, so model your consumption, not just your headcount.
  • Both, which is most engineering organizations. Run them as two accounts on purpose, each with a named owner. Then decide explicitly whether heavy Claude Code work is funded from plan limits or from Console credits, because both routes exist and only one of them is budgeted.

That last decision is the one teams skip. With OpenAI the wall makes it for you; with Anthropic the doors mean somebody has to choose - and it is much cheaper to choose deliberately in advance than to discover the answer on an invoice.

References