Claude Team Plan: What You Get and How to Set It Up

The Claude Team plan puts a group of people, from two up to 150 seats, in one Claude organisation that an admin controls. Each person gets their own seat, with Claude Code and Cowork included, and the organisation gets project sharing, admin-managed connectors, single sign-on and a commercial default of no model training on your content. It is the step up from everyone buying Pro separately, and the step below Enterprise, which adds audit logs, SCIM, custom roles and compliance tooling.

What the Team plan includes

Anthropic’s page on what the Team plan is↗ (opens in a new tab) lists these, and they are the parts worth knowing before you buy:

• Two kinds of seat. Standard seats have more usage per session than Pro; Premium seats have several times more again, for heavy users. You can mix them, and usage limits apply per member, not to the team as a pool.
• Claude Code on every seat, used from the terminal, IDE, desktop app or web by signing in with the organisation’s account.
• Cowork, on by default for the organisation, with an owner switch to turn it off.
• Projects that can be shared with the whole organisation or with invited members. How that sharing works is covered in can Claude Projects be shared?.
• Connectors to Google Drive, Gmail, Google Calendar, GitHub, Microsoft 365 and Slack, plus custom connectors to any remote MCP server that an owner adds for the organisation.
• Enterprise search: a ready-made search project provisioned for every member.
• Admin features: single sign-on and domain capture, just-in-time provisioning, role-based permissions, and spending caps for the organisation and for each member.

Data use is the other reason teams move. Anthropic’s privacy page on whether your data is used for training↗ (opens in a new tab) says inputs and outputs from commercial products, which include Team and Enterprise, are not used to train models by default, unless you report feedback or a bug. On Free, Pro and Max it is a setting each person chooses.

Two things are not on Team today. Computer use in Cowork and in Claude Code in the desktop app is a Pro and Max beta, and Anthropic says Team and Enterprise do not have it yet. Dispatch, which sends a task from your phone to the desktop app, is also Pro and Max only.

Team vs Pro vs Enterprise

Feature by feature, as Anthropic documents them today. Prices are left out; check the plan pages for those.

• People. Pro: one person. Team: two to 150 seats, Standard and Premium mixed. Enterprise: from 20 seats bought online (50 through sales), so it overlaps Team’s range and the choice turns on controls, not headcount.
• Claude Code and Cowork. All three plans. On Enterprise, some Cowork surfaces depend on the admin enabling them.
• Sharing projects. Pro: no. Team and Enterprise: yes, organisation-wide or with invited members.
• Central billing and member management. Team and Enterprise.
• Single sign-on. Team and Enterprise.
• Connector controls. Team and Enterprise: owners add and govern connectors for the organisation.
• Training on your content. Pro: your choice, in a privacy setting. Team and Enterprise: off by default.
• Audit logs, SCIM, custom data retention, the Compliance API and IP allowlisting. Enterprise only.
• Custom roles. Enterprise only; Team has the built-in roles.
• Computer use and Dispatch. Pro and Max only.

The honest summary: if you need one person to have more usage, Max is the answer, not Team. If you need several people to share projects and connectors under one set of rules, Team is. If your security team asks for audit logs or SCIM, the answer is Enterprise, described on Anthropic’s page on what the Enterprise plan is↗ (opens in a new tab).

How to set up a Claude team

The order below follows Anthropic’s guide to getting started with the Team plan↗ (opens in a new tab). Menu names move, so look for the section rather than an exact button.

1. Create the organisation. Sign in at claude.ai with a work email on your company’s domain, not a personal address, and choose the Team plan during onboarding. The organisation has exactly one Primary Owner, which can later be transferred.
2. Choose seats. Decide how many Standard and how many Premium seats you need. Start with Standard for most people and add Premium for the two or three who will live in Claude Code or Cowork all day.
3. Invite people. In Organization settings, Members, use Add member for one person, Bulk add for a list, or an invite link from the organisation and access settings. Invitees need an email on an allowed domain, and you pick each person’s seat type and role as you add them.
4. Set roles. The built-in roles are User, Admin, Owner and Primary Owner. Admins manage members and features; only Owners and the Primary Owner reach billing and can make new Admins or Owners. Keep Owners to two people.
5. Turn on single sign-on if you have an identity provider. Anthropic’s setup guide names Google Workspace, Microsoft Entra ID, Okta, OneLogin and Ping Identity, and Team lists domain capture alongside single sign-on.
6. Add connectors. An Owner opens Organization settings, Connectors, and adds the directory connectors you want or a custom connector by its URL. Each member then connects under Customize, Connectors, signing in to the service as themselves.
7. Review Cowork. In Organization settings, Cowork, decide whether cloud sessions, the built-in browser and the automatic approval mode are allowed. Leave “Always allow” for write-capable connector tools off until you know which tools people use.
8. Point developers at Claude Code. They run claude, choose the Claude account with subscription, and authorise the organisation.

Anthropic’s page on roles and permissions↗ (opens in a new tab) has the full matrix if your admin wants to check a particular permission before handing it out.

Good habits for the first week

• Make one shared project per team, not per person, with one owner for its instructions. What works in a shared project and what does not is in Claude Projects for teams.
• Connect only the connectors a real task needs this week. Each is a door into a system, and it is easier to add one later than to explain why it was open.
• Agree where results go. Chats and Cowork sessions are private to the person who ran them, so a finding that matters to the team has to land somewhere the team can see.
• Write down two or three house rules, such as “never send email without a person reading it”, and put them in the shared project’s instructions.
• Check usage at the end of the week and move seats between Standard and Premium based on what people actually used.

Where the team’s work list lives

A Team plan shares knowledge and tools, not a record of who is doing what. That is a board’s job. fenbs is a small task board where people and AI assistants are both members: Claude joins as a custom connector, over MCP, and acts with the role of the person who connected it, narrowed by the scopes they ticked. Roles are defined per company, with Client and Reporter suggested for outsiders, and every change is recorded in History with who made it, including “Claude via” the person behind it. Owners who add fenbs as an organisation connector give every member the same board to report back to.