Seven paths to Claude

Both are seat plans for people, sold by Anthropic, with the same apps: Claude on web, desktop and mobile, Claude Code, Cowork, Design, connectors. The difference is governance depth, team size and the billing model. Team is for 2 to 150 people, bought with a card, with a fixed price per seat and usage included per member (Standard or Premium seats); it has SSO, domain capture, just-in-time provisioning, role-based permissions and spend caps.

Enterprise starts at 20 seats self-serve or 50 seats through sales, charges an access fee per seat and then meters all usage at API rates with spend limits per organization, group and user, and adds SCIM provisioning, audit logs, a Compliance API, custom data retention, IP allowlisting, customer-managed encryption keys, a HIPAA-ready configuration and a US-only inference option. An Enterprise parent organization can also link Claude Console (API) organizations, which is how companies keep people and software under one contract.

Rule of thumb: pick Team if you are under 150 people and your security policy is satisfied by SSO and role-based access; pick Enterprise if you need SCIM, audit trails or retention controls, if a large share of the team will use Claude Code or Cowork heavily, or if you will exceed 150 seats. Moving from Team to Enterprise is a one-way migration, so settle the governance question early.

All five developer platforms serve the same models with the same Messages API shape, prompt caching, tool use and extended thinking. They differ in who operates inference, how you pay, which server-side features exist, and where inference can run.

Only two paths keep model inference inside the EU today: Amazon Bedrock (EU inference profiles and in-region endpoints such as Stockholm and Ireland) and Google Cloud (the eu multi-region endpoint). The Claude API, Claude Platform on AWS and Microsoft Foundry route globally by default and can be pinned to the US only; Foundry's EU data zone is listed as not available; the Claude apps offer a US-only option on Enterprise and no EU option. Data residency is not the same as GDPR compliance — Anthropic-operated paths come with a data processing addendum and transfer safeguards — but where a policy demands EU processing, the list is two.

SSO is available on Claude Team, Claude Enterprise and Claude Console organizations after domain verification; SCIM is available on Enterprise and Console organizations only, so a Team plan gives you single sign-on with just-in-time provisioning but no automated deprovisioning. On the cloud paths identity is the cloud's own — AWS IAM, Google Cloud IAM, Microsoft Entra ID — and the audit trail lives in CloudTrail, Cloud Audit Logs or Azure Monitor rather than in Anthropic's Compliance API.

No path trains on your data by default. Zero data retention is available on request on the Claude API and Claude Platform on AWS, and as an option on Microsoft Foundry, with the newest Fable and Mythos models requiring 30-day retention; on Bedrock and Google Cloud, retention is governed by AWS and Google because Anthropic never receives the content. HIPAA-ready configurations exist on the Claude API and Claude Enterprise (self-serve BAA) and through AWS on Bedrock — not on Claude Platform on AWS or Foundry.

The Console is not a separate product. Every Claude API organization has one: it is where keys, workspaces, credits, usage and rate limits live, and where the playground replaced the old Workbench in August 2026. You do not buy it; you use it.

The confusion comes from claude.ai: a Team or Enterprise subscription buys seats for people, while the Console holds credits for software. They are billed separately, and a Claude Team seat does not give your developers API credits — nor does an API balance give your staff the Claude apps. Enterprise customers can link the two under one parent organization for shared governance, but the meters stay distinct.

Yes, and most companies end up on two. The typical sequence for a Greek or EU SME: Claude Team seats for the people who write, analyze and sell; a Claude API organization (through the Console) for the first automation or product feature; then, once governance requirements or headcount grow, Claude Enterprise as the parent organization that links the Console org — or a cloud path if procurement moves the API spend onto AWS, Google Cloud or Azure.

Two things are one-way or sticky: Team → Enterprise cannot be reversed, and a Claude Platform on AWS organization is always new — an existing Claude API organization cannot be converted. Model IDs, SDKs and prompts port across all five API platforms, so switching hosts is an engineering task, not a rewrite; only the features that exist solely on Anthropic-operated paths (server-side tools, Files, Batches, Skills, Managed Agents) stay behind when you move to Bedrock or Google Cloud.

Read down until a rule matches your situation. These are the rules we apply on a first call, in the same order — team size and governance first, then residency and compliance, then what you are building and where your cloud spend lives.