How to Set Up Claude: Team and Enterprise Guide

If you are considering or already have signed up for Claude for your organization, this guide is for you. Specifically, we will look at the Team and Enterprise plan, covering the admin settings and connector choices worth reviewing first.

This guide assumes you’ve already committed to either the Team or Enterprise plan and are preparing to roll out the tool  to everyone. Most of the settings live in one place, under Organization settings in Claude’s admin console. Only users at the Owner level can change many of them, so whoever holds that role should be the one working through this list.

Why does it matter? A business AI subscription behaves more like shared infrastructure than a typical app. It can reach your files, your inbox, and eventually your client data (if you let it). Anthropic’s defaults are sensible for a generic organization, but they can’t account for your policies, your compliance requirements, or the kind of data your team handles.

One caveat before we start: Claude’s admin settings change often. Menus get renamed, features move out of beta, and some controls only exist on Enterprise. What follows reflects Anthropic’s help center as of October 2026.

Who Holds the Keys?

Every Team and Enterprise organization has four built-in roles: User, Admin, Owner, and Primary Owner. The differences matter more than the names suggest. The big takeaways from Anthropic’s roles and permissions guide:

• Admins can invite and remove members
• Owners and the Primary Owner can turn on connectors and capabilities, invite new users and buy new seats
• Only the Primary Owner request a full data export

There’s only one Primary Owner, and it’s often whoever signed up for the plan. Check who holds it today and make sure it can be transferred easily if that person changes roles. Anthropic allows the Primary Owner to be a service account that isn’t tied to one person, which is worth considering but not recommended. We also recommend keeping the Owner list short and limited to the people responsible for your systems.

Control How People Sign In

Single sign-on (SSO) is available on both Team and Enterprise plans, and Anthropic’s SSO setup guide walks through the technical setup. The first step is verifying your email domain with a DNS record. Once that’s done, two settings deserve a hard look.

Require SSO for Claude is the first one to understand. With this requirement turned on, everyone logs in through your identity provider, whether that’s Microsoft Entra, Google, or Okta. When someone leaves, disabling their account in one place cuts off Claude too, giving you one less login to worry about.

The second setting to consider is Restrict organization creation. It stops people from creating new Claude organizations, including personal accounts, on your company domain. This matters because personal plans run under consumer terms, which handle data differently than your business plan. You may think that your employees wouldn’t be bold enough to create a personal account under your company domain, but it happens more often than you think.

The same guide includes a warning worth repeating: If SSO is required and your identity provider’s signing certificate expires, nobody can sign in to fix it. You would need to contact Anthropic support to regain access. Put the certificate’s expiration date on your IT calendar so it gets renewed ahead of time.

Finally, outside of the settings, be sure to make an organization decision regarding how people get added. The default is invite-only. Just-in-time provisioning, available on both plans, creates an account the first time someone signs in through SSO. On Enterprise, SCIM directory sync goes further and removes people automatically when they’re removed from your identity provider. The details are in Anthropic’s provisioning guide.

Know Where Your Data Goes

This is the first question most leaders ask, so let’s tackle it head-on.  Anthropic’s privacy center says it will not use inputs or outputs from commercial products to train its models by default, and that includes Team and Enterprise. However, there’s one exception worth knowing.

When someone clicks the thumbs up or thumbs down button during chats, the whole conversation goes to Anthropic as feedback. It can be stored for up to five years and may be used for training. Owners can turn this off by toggling off Rate Chats. You can get to it through your organization’s settings:

Organization Settings > Data and Privacy > Rate Chats

If your team works with student records, resident data, client financials, or any proprietary data, we recommend turning it off.

While you’re on that page, look at the settings for project sharing. Projects are shared workspaces with their own files and instructions. Users can share them across the organization, and Public Projects lets anyone in the company see and chat in them. Both are on by default, per Anthropic’s project sharing guide. If your teams regularly work with sensitive files, consider turning public projects off and letting people share with specific colleagues instead.

Offboarding is worth planning for too. Anthropic’s article on removed users explains that when someone is removed, other members lose access to their chats and to any of their projects that were set to private. Add a step to your offboarding checklist to move or reshare important Claude projects before an account is removed.

Decide What Claude Is Allowed to Do

Under Organization settings > Capabilities, you’ll find Code execution and file creation. It’s on by default for Team plans and new Enterprise organizations, according to Anthropic’s file creation guide. It’s what lets Claude build spreadsheets, documents, and charts. Most teams will want it. The setting to check underneath it is network access. That controls whether Claude’s workspace can reach the internet while it works. Tighter is safer, so open it up only when someone has a real need.

Another check under your organization’s settings is usage. If you turn on usage credits, people can keep working after they hit their seat’s limit, billed at API rates. Anthropic’s usage credits guide shows how to set a monthly cap for the whole organization and individual people. We recommend setting a cap at the same time you turn usage credits on.

Connectors Are Where to Slow Down

Connectors are what make Claude feel useful to most people. They let it search Google Drive, read email, look up tickets, or pull from Microsoft 365. They’re also where most of the real risk lives, so this is where we spend the most time.

The mechanics are simple. An Owner adds a connector in the organization’s settings. That makes it available, but nobody gets access automatically. Each person still must sign in to the service themselves. Anthropic’s connectors guide says Claude inherits each person’s permissions from the connected service. If an employee can’t open a file in SharePoint, Claude can’t open it for them either.

This is a helpful safeguard, and it also means Claude’s reach depends on how well your existing permissions are managed. Anything shared broadly in your file systems can be found through a connector by anyone with that access. Reviewing sharing permissions in your file systems is a good step to take before connectors go live.

Next, set tool permissions. Owners can limit what each connector is allowed to do across the whole organization. Each action can be set to Always allow, Needs approval, or Blocked, and individual users can’t override it.

So how does this work in practice? Anthropic’s own examples are practical:

• Let Claude search and summarize email, but block sending.
• Let it read Drive files, but block editing.

Our usual starting point is simple. Reading is allowed, writing needs approval, and deleting is blocked. You can loosen things once people trust the workflow.

Be more careful with custom connectors. These connect Claude to services Anthropic hasn’t reviewed. Anthropic’s custom connector guide warns that a malicious server can hide instructions meant to make Claude take actions nobody asked for, called prompt injection. The guide also notes that a connector set up with a shared API key gives every user that key’s access, which undoes the per-person permissions described above. Only add custom connectors from vendors you already trust, and use a credential limited to what people need.

Two more details are worth knowing:

• Connected services process data under their own terms, which may include servers outside the United States.
• Connectors only work in private projects, and chats with synced content can’t be shared.

Neither is a problem, but both are worth explaining to users during rollout so they understand how connectors work. If you’re not sure where to start, pick one or two connectors tied to a real job. The file system people already search all day is usually the right first choice. Roll it out to a small group, watch how they use it, then expand as you see fit.

If You’re on Enterprise

Enterprise adds controls that regulated organizations will want. Those include:

• Audit logs
• Custom data retention 
• Custom roles for departments
• IP allowlisting

Owners can also stop company services from being connected to personal Claude accounts. For schools and municipalities with records retention rules, the retention and audit settings deserve a conversation with whoever owns compliance.

Where to Go From Here

Most organizations can work through these settings in an afternoon. The tool works out of the box, so it’s easy to overlook. Taking the time now means the defaults reflect your organization’s decisions.

At Macro Connect, we help organizations set up tools like Claude so they’re useful and secure from the start. That usually begins with identity, permissions, and a clear policy on what data goes where. If your organization just rolled out Claude, or is about to, we’d be glad to walk through this process with you.