Claude Enterprise: Features, Pricing, and How to Roll It Out

Claude Enterprise is Anthropic’s top business plan for Claude. As of October 2026 it costs US$20 per seat per month, billed annually, plus usage that scales with model and task. It includes everything in the Team plan, then adds SCIM, audit logs, a Compliance API, custom data retention, IP allowlisting and a HIPAA-ready option. Each seat covers Claude on web, desktop and mobile, plus Claude Code and Cowork, while every token is billed at standard API rates.

This guide is for IT leads, security reviewers and the people who will train staff. It explains what the plan includes, how the cost works, how it differs from Team, and a rollout checklist you can copy. All plan facts below come from Anthropic’s own pricing and help pages, read on 13 October 2026. Plans change often, so verify current pricing on the official site before you sign.

Related reading: Claude for Teams, Claude pricing, and Claude for small business. Key terms used in this guide: prompt injection and context window.

Claude Enterprise vs Team at a glance

Most buyers are really choosing between Team and Enterprise. The table compares the two plans using the official pricing page and the Enterprise plan overview.

The key difference is the billing model. A Team seat bundles a usage allowance. An Enterprise seat buys access and controls, and the Enterprise overview says there are “no per-seat usage limits and no included token allowance.”

Key features of Claude Enterprise

Enterprise adds controls that matter once Claude touches company data at scale. Here is what each group of features does in plain terms.

Identity and access: SSO, SCIM and roles

Single sign-on is available on Team and Enterprise plans. Setting it up needs an Owner or Primary Owner role and access to your company’s DNS settings to verify your email domain. The same guide warns that enforcing SSO can lock out people who are not assigned to the Anthropic app in your identity provider.

SCIM is the Enterprise piece. JIT provisioning works on Team and Enterprise, but SCIM is limited to Enterprise and Console organizations. SCIM syncs your directory, so new hires get access and leavers lose it without manual steps. Advanced group mappings let you decide which role each group receives. The pricing page also lists role-based access with fine grained permissioning for Enterprise.

Audit logs and the Compliance API

Audit logs are available for Enterprise organizations only. Owners and Primary Owners can export them, and one export covers the past 180 days. The download link stays active for 24 hours. Logged events include sign-ins, invitations, user removal, SSO and domain changes, file uploads, and conversation creation or deletion. Chat titles and content are not in the export, only their IDs.

For deeper monitoring there is the Compliance API. It gives programmatic access to activity feed events, chat data and file content, so your security tools can pull records on a schedule. Only the Primary Owner can switch it on. The same page notes it is not offered to Public Sector organizations, and some surfaces, such as cloud sessions in Claude Code, are not covered.

Data retention, encryption and US-only inference

By default, Enterprise data is kept indefinitely. Owners can set a custom retention period with a minimum of 30 days. The Enterprise overview also lists customer-managed encryption keys and US-only inference. Note the cost side: with US-only inference on, usage is billed at 1.1x standard API rates for Claude Opus 4.6, Sonnet 4.6 and later models.

On training, the pricing page states “no model training on your content by default” for the Team plan, and Enterprise includes all Team features. Network-level access control and IP allowlisting let IT block access from outside approved networks. IP allowlisting is set up via Anthropic support; list every office and VPN range first.

Connectors, Claude Code and context limits

Connectors let Claude pull context from work apps. The Enterprise overview names Google Drive, Gmail, Google Calendar, GitHub, Microsoft 365 and Slack. The pricing page lists admin controls for remote and local connectors, so IT decides which ones staff can use.

Claude Code is part of the Enterprise seat, but its usage draws on the same API-rate billing as chat.

Context size is not an Enterprise perk anymore. Anthropic’s context window article applies the same model table to Pro, Max, Team and Enterprise. Several current models, such as Claude Opus 5.5 and Sonnet 5.5, list a 1M token window in chat, and other models default to 200K. Check the live table, since it changes with each model release.

Claude Enterprise pricing: how the cost works

Your bill has two parts. The first is a fixed seat fee. The second is usage, charged on actual token consumption at standard API rates. Heavier models and longer tasks cost more, which is why the pricing page says usage cost “scales with model and task.”

Self-serve vs sales-assisted

The Enterprise overview describes two ways to buy, with the same features and seat price:

• Self-serve: bought online, minimum 20 seats, card or ACH, USD only. Usage credits are bought up front, and when they run out, usage stops until someone with billing rights buys more.
• Sales-assisted: arranged with Anthropic’s sales team, minimum 50 seats. It supports invoicing and several currencies, and usage is billed monthly in arrears.
• AWS Marketplace: Enterprise is also listed there for companies that buy software through AWS (sales-assisted only).

The pricing page adds that sales-assisted buyers can get consultation, trials and customer success support at certain spend levels. Ask the sales team what applies to your size.

Spend limits and a worked example

Owners can set spend limits for the whole organization and for each user. A user is capped by whichever limit is lower. You can raise limits to unlimited, but you cannot switch off usage billing.

Hypothetical example: a 60-person firm buys self-serve Enterprise. Seat fees alone come to 60 x $20 x 12 = $14,400 a year, based on the list price above. Usage is on top. Ten developers running Claude Code all day will burn far more tokens than 50 staff drafting emails. A sensible setup gives developers a higher per-user limit, sets a modest default for everyone else, and reviews spend after the first month. Check current API rates and limits on the official site before you build a budget.

Who should consider Claude Enterprise?

Enterprise fits when control matters more than simplicity. Use these signals:

• You need automatic user lifecycle. If HR changes must reach Claude access without tickets, you need SCIM.
• Auditors will ask who did what. Audit logs and the Compliance API are Enterprise only.
• You are a HIPAA covered entity or business associate and will process protected health information (PHI) in Claude. The HIPAA-ready configuration is for Enterprise plans only, on both self-serve and sales-assisted, and the Primary Owner accepts a BAA in settings.
• You must limit how long chats exist. Custom retention is an Enterprise control.
• You are above 150 people. Team tops out at 150 seats on the pricing page.

Team can be enough for a smaller group who want SSO, shared projects and predictable seat costs, with no audit or retention demands. Many firms start on Team, learn how staff use Claude, then move up.

For the trust review, Anthropic’s certifications page lists SOC 2 Type I and Type II, ISO 27001:2022, ISO/IEC 42001:2023 and a HIPAA-ready configuration with a BAA for its commercial products. Request the reports through Anthropic’s Trust Portal rather than relying on summaries.

What to know before deciding: limitations and common mistakes

Enterprise solves governance problems, but it adds new ones to manage.

• Usage has no ceiling unless you set one. With no included allowance, a busy month costs more. Set limits on day one.
• HIPAA mode changes features. Turning on HIPAA does not cover every feature. Process patient data only in features the Implementation Guide lists as covered. Claude Code and Cowork are covered only after the Primary Owner also applies the HIPAA configuration to their local modes; then cloud features such as Claude Code on the web are switched off. Enabling HIPAA resets some settings and cannot be reversed from settings.
• Audit exports leave out content. If reviewers need chat text, plan for the Compliance API instead.
• SSO enforcement can lock people out. Test with a pilot group before enforcing it for all.
• Connectors widen the attack surface. When Claude reads outside emails, files or web pages, hidden text can try to steer it. This is called prompt injection. Grant the least access each team needs, keep human approval on for sending, deleting and payments, and decline connectors that ask to read every file in a drive.
• Staff still paste the wrong things. Even on Enterprise, tell people never to paste passwords, card numbers or ID documents, and never patient details unless your approved policy covers that use, and only on a HIPAA-enabled organization, in features covered by your BAA.

Decision framework and rollout checklist for IT and training staff

Use this four-phase checklist. It turns the plan’s features into tasks with owners.

Phase 1: Decide and buy

1. List the controls you must have (SCIM, audit logs, retention, HIPAA). If none apply, price Team first.
2. Estimate seats. 20 to 49 seats points to self-serve; under 20, price Team; 50 or more opens sales-assisted invoicing.
3. Pick the people who will hold the Primary Owner and Owner roles. Several key settings need them.

Phase 2: Configure identity and data

1. Verify your domain and set up SSO with your identity provider.
2. Turn on SCIM and map groups to roles.
3. Set organization and user spend limits.
4. Choose a retention period (shortening it schedules older data for permanent deletion as soon as you save; projects keep their own retention) and decide on US-only inference and the Compliance API.
5. Enable only the connectors each team needs.

Phase 3: Pilot

1. Start with 15 to 30 people from two or three teams.
2. Run for four weeks, then export audit logs and review usage and spend per user.
3. Collect the tasks that worked and the ones that did not.

Phase 4: Train and scale

1. Publish a one-page acceptable-use policy: what data is allowed, which connectors are on, who to ask.
2. Give each department three tested prompt examples from the pilot.
3. Teach staff to check outputs against sources before sending anything outside the company.
4. Expand in waves and review spend after each wave.

Training staff who are new to AI

The rollout fails if people do not know how to use the tool well. Training staff should plan for mixed comfort levels, especially among employees meeting AI for the first time. Short practice sessions on real tasks work better than one long demo.

Coursiv offers guided, bite-sized AI lessons people can take at their own pace. Pair lessons like these with your internal policy so staff learn both the skill and your company’s rules.

Next step

Write down the three controls your security team will ask about, then check each one against the table above. If Team covers them, start there. If not, run the Phase 1 checklist and book a call with Anthropic’s sales team. While IT sets up the plan, help your staff build the basics so they are ready on launch day: Explore Coursiv AI lessons.

Related guides: is it safe to use AI tools at work and Claude usage limits.

FAQ