- AI tools
- Artificial Intelligence
- Business
- Business Intelligence
Claude Enterprise: Pricing, Features, Security, and How It Compares
Claude Enterprise costs $20 per seat each month plus usage. Learn what it includes, how security works, and how it compares with Claude Team, ChatGPT, and Grok.
Claude Enterprise combines Anthropic’s AI workspace, Claude Code, Cowork, connected company knowledge, and enterprise controls. Its pricing is unusually important to understand: the seat opens the platform, while usage is billed separately.
Reviewed September 7, 2026
Anthropic’s current public pricing lists Claude Enterprise at $20 per seat per month, billed annually, plus usage at standard API rates. That is not an all-inclusive $20 subscription. This guide explains the full model, including minimum seats, security controls, Claude Code, Cowork, connectors, and the difference from Claude Team.
| Question | Short answer |
|---|---|
| What is Claude Enterprise? | Anthropic’s managed AI plan for organizations that need scalable usage, advanced governance, and compliance controls. |
| How much does it cost? | $20 per seat per month on an annual term, plus usage billed at API rates. Taxes and negotiated terms may vary. |
| What is the minimum? | 20 seats for self-serve Enterprise and 50 for sales-assisted Enterprise, according to Anthropic’s current help center. |
| Is business data used for training? | No. Anthropic says it does not train generative models on Claude for Work customer data. |
| Why choose it over Team? | SCIM, audit logs, Compliance and Analytics APIs, custom retention, finer roles, network controls, CMK, US-only inference, and HIPAA-ready options. |
What is Claude Enterprise?
Claude Enterprise is the large-organization tier of Claude, Anthropic’s AI assistant and work platform. It gives employees access to Claude across web, desktop, and mobile, and includes Claude Code for software work and Claude Cowork for broader agentic tasks. The organization receives centralized billing, identity management, permissions, usage controls, and compliance tooling.
Claude Enterprise is a plan, not a distinct model. Users may work with different Claude model families depending on the task and the organization’s settings. Administrators can control which models are available, choose defaults, and set spending limits. This matters because model choice changes both capability and cost.
Anthropic’s current Enterprise page positions the product as a common layer for knowledge work, coding, data analysis, and connected company information. The more practical definition is simpler: it is Claude with the deployment controls a large IT, security, legal, and finance organization needs.
Claude Enterprise pricing in 2026
Claude Enterprise uses two charges:
- A seat fee: Anthropic’s current pricing page lists $20 per seat per month, billed annually.
- Usage: tokens consumed across Claude chat, Claude Code, and Cowork are billed separately at the standard API rate for the model and task.
This is the most important fact for a buyer. The $20 seat is an access and governance fee, not a bundle of unlimited Claude usage. A team that mostly asks short questions may have a modest consumption bill. A team running long coding sessions, large-document analysis, or autonomous Cowork tasks can spend considerably more.
Self-serve vs. sales-assisted Enterprise
Anthropic offers two purchase paths with the same core seat model:
| Area | Self-serve Enterprise | Sales-assisted Enterprise |
|---|---|---|
| Minimum seats | 20 | 50 |
| Seat billing | Annual | Annual |
| Usage payment | Credits purchased in advance | Monthly usage billed in arrears |
| Payment and currency | USD; card, debit card, or ACH | Multi-currency and invoicing options |
| Best fit | Organizations that can buy and manage the plan online | Organizations needing procurement support, tailored terms, or dedicated success |
Anthropic’s billing guide says new Enterprise contracts use this single-seat, usage-based structure. Some existing customers remain on legacy seat types until renewal, which explains why older descriptions online may not match a new quote.
How to estimate the real cost
Start with the annual seat commitment, then model consumption by role. A software engineer using Claude Code all day does not resemble an executive asking for a few summaries. Segment expected users into light, regular, and power-user groups; test representative tasks; record token cost; and add a buffer for adoption.
Enterprise admins can set organization and individual spend limits. Use those controls during a pilot so one experimental workflow cannot consume the entire shared budget. Anthropic also advises using a general-purpose model for routine work and reserving premium models and high effort levels for the tasks that justify them.
What does Claude Enterprise include?
Claude chat for everyday knowledge work
Employees can draft, summarize, research, analyze files and images, create structured outputs, and work through complex questions. Projects and shared organizational context help teams keep related conversations and source material together.
Claude Code
Claude Code brings the assistant into terminals and supported development environments. Developers can ask it to understand a codebase, plan and implement changes, run tests, investigate failures, review code, and write documentation. Enterprise admins can manage model access, policies, permissions, and spend across coding users.
Claude Cowork
Cowork extends the agent-style workflow beyond code. It can handle longer, multi-step tasks that involve files and workplace tools, with the user reviewing progress and outputs. This can be useful for preparing analysis, organizing information, or producing deliverables, but high-impact actions still need clear permission and human-review rules.
Connectors and enterprise search
Claude can connect to workplace systems including Google Drive, Gmail, Google Calendar, GitHub, Microsoft 365, and Slack. That lets an employee retrieve relevant documents, messages, code, and schedules without repeatedly uploading files. Connector access must still be configured and governed; a model should not become a shortcut around source-system permissions.
Larger working context
Anthropic’s current plan comparison lists a 500K context window on the default Enterprise model, compared with 200K on Team. In plain English, Enterprise can work with more material in one active task. Context size is not the same as perfect recall, and it does not remove the need for good retrieval, document structure, or evaluation.
Administration and analytics
Central membership, role-based permissions, spend controls, usage analytics, audit logs, and programmatic Analytics and Compliance APIs help organizations operate Claude as a managed service instead of a collection of individual subscriptions.
How secure is Claude Enterprise?
Claude Enterprise adds controls for identity, data handling, monitoring, and regulated use. The correct question is not simply “Is Claude secure?” It is whether the exact configuration, contract, integrations, and workflow meet your organization’s requirements.
No model training on commercial customer data
Anthropic states that it acts as a processor for Claude for Work customer data and does not use that data to train generative models. This differs from consumer-product settings and is one reason employees should work inside the approved organizational workspace.
Identity and access management
Enterprise includes SSO, domain controls, SCIM, audit logs, and fine-grained role-based access. Administrators can manage model availability, connectors, Claude Code policies, organization-wide settings, and who can use higher-risk or higher-cost capabilities.
Retention, monitoring, and encryption controls
Custom retention policies let eligible organizations decide how long content remains available. The Compliance API can expose activity logs, chat histories, and file content for approved monitoring and governance workflows, while the Analytics API provides aggregate adoption and usage data. Customer-managed encryption keys can add control over the keys protecting chats, projects, and files.
Network and location controls
Anthropic lists network-level controls and IP allowlisting for Enterprise. Usage-based Enterprise customers can also enable US-only inference. That setting applies to model processing, not automatically to data storage or third-party connectors, and currently adds a premium to eligible model usage.
HIPAA-ready configuration
Eligible organizations can activate a HIPAA-ready Enterprise configuration and accept Anthropic’s BAA. This is not automatic: the Primary Owner must enable the appropriate settings, and only covered features and workflows should handle protected health information. Confirm the current coverage with Anthropic and your own compliance team.
Claude Team vs. Claude Enterprise
| Area | Claude Team | Claude Enterprise |
|---|---|---|
| Team size | 2 to 150 users | 20-seat self-serve or 50-seat sales-assisted minimum |
| Standard pricing | $20 annually or $25 monthly per seat; included usage limits | $20 per seat monthly on an annual term, plus API-rate usage |
| Claude Code and Cowork | Included; plan usage limits apply | Included; consumption is billed |
| SSO and basic administration | Yes | Yes |
| SCIM | No | Yes |
| Audit logs and Compliance API | No | Yes |
| Custom retention and fine-grained roles | Limited | Yes |
| Context on default model | 200K | 500K |
| CMK, IP allowlisting, US-only inference | No | Available |
| HIPAA-ready offering | No | Available for eligible configurations |
The surprising part is pricing. Team’s standard annual seat and Enterprise’s access seat both start at $20, but Team includes a usage allowance while Enterprise adds metered usage. Enterprise is therefore not an automatic bargain at scale. Its justification is the governance and capacity model, not the seat price alone.
Claude Enterprise vs. ChatGPT Enterprise vs. Grok Enterprise
All three products provide a managed AI workspace and promise not to train on business data by default. Their emphasis differs.
| Platform | Distinctive fit | Important buying question |
|---|---|---|
| Claude Enterprise | Long-form knowledge work, coding with Claude Code, agentic work through Cowork, and granular consumption controls | Can you forecast and govern usage-based spend across roles? |
| ChatGPT Enterprise | Broad employee productivity suite, mature apps and company knowledge, analytics, agents, and enterprise administration | Do its workspace and connector controls fit your data estate and procurement needs? |
| Grok Enterprise | Real-time web and X search, multimodal creation, connected tools, and an optional isolated Vault architecture | Does the newer enterprise ecosystem meet your governance and reliability bar? |
Do not choose solely from a public benchmark. Run the same representative tasks in each platform and score answer quality, citation quality, completion rate, latency, admin effort, data controls, and total cost.
Who should consider Claude Enterprise?
Claude Enterprise is particularly relevant when:
- developers already rely on Claude Code and the company wants centralized policy and spend;
- knowledge workers handle large collections of documents or complex, multi-step work;
- security requires SCIM, audit logs, custom retention, IP restrictions, or customer-managed keys;
- compliance teams need programmatic access to activity and content records;
- the organization prefers usage-based capacity to per-user message limits;
- a HIPAA-ready configuration or US-only inference is part of the requirement.
Claude Team may be the better starting point when the organization has fewer than 150 people, included usage is easier to budget, and Enterprise-only controls are not yet required.
Tradeoffs to understand before buying
Usage-based billing can be efficient or unpredictable
Paying for actual consumption can align cost with value, but heavy Claude Code and Cowork usage can grow quickly. A pilot should measure cost by workflow and role, not just average tokens per employee.
Long context does not replace retrieval design
A larger context window helps with big files and extended tasks, but stuffing every available document into one conversation can reduce clarity and raise cost. Good source selection, permissions, chunking, citations, and evaluation still matter.
Agents raise the stakes
Claude Code and Cowork can do more than draft text. Any capability that edits files, calls tools, or acts across systems needs least-privilege access, approval points, logging, and rollback plans.
Feature labels move quickly
Claude’s models, products, limits, and plan packaging evolve often. Put required capabilities and data terms in the order form rather than relying on a marketing page captured during evaluation.
A practical Claude Enterprise evaluation checklist
- Select 10 to 20 real workflows across chat, documents, code, and connected sources.
- Classify the data and remove any workflow the proposed configuration cannot support.
- Measure quality, citations, task completion, human correction time, and consumption cost.
- Test SSO, SCIM, roles, connector permissions, offboarding, exports, and audit visibility.
- Set model defaults and spend limits before expanding the pilot.
- Review retention, inference location, BAA coverage, subprocessors, and incident terms.
- Compare the full annual cost with Claude Team and at least one competing enterprise assistant.
Frequently asked questions
How much does Claude Enterprise cost?
Anthropic currently lists Claude Enterprise at $20 per seat per month, billed annually, plus usage at standard API rates. Taxes, currency, discounts, support, and negotiated commitments can change the final contract.
Does Claude Enterprise include Claude Code?
Yes. The current single Enterprise seat includes Claude on web, desktop, and mobile, plus Claude Code and Cowork. Usage across those surfaces is metered and billed to the organization.
What is the Claude Enterprise minimum seat count?
Anthropic’s current help center lists a 20-seat minimum for self-serve Enterprise and a 50-seat minimum for sales-assisted Enterprise. Claude Team supports 2 to 150 users.
Does Anthropic train on Claude Enterprise data?
No. Anthropic says it does not use Claude for Work customer data to train its generative models. The organization remains responsible for its connected tools, user permissions, and internal data policy.
Is Claude Enterprise unlimited?
Current usage-based Enterprise plans do not use a fixed per-seat message allowance. Usage is metered and billed at API rates, and administrators can set organization and user spend limits. That is different from free, Pro, Team, and legacy seat-based limits.
Is Claude Enterprise HIPAA compliant?
Anthropic offers a HIPAA-ready Enterprise configuration for eligible organizations, but BAA coverage is not automatic. An authorized owner must activate the configuration and accept the BAA, and only covered features should be used with protected health information.
Is Claude Enterprise better than ChatGPT Enterprise?
Neither is universally better. Claude may be a strong fit for long-form reasoning, coding, and usage-metered agent work; ChatGPT may fit organizations that prefer its wider productivity ecosystem and OpenAI administration stack. Test your own workflows, controls, and costs.
Sources and methodology
This article was researched from Anthropic’s public product and support documentation and reviewed on September 7, 2026.
- Claude plans and pricing
- Anthropic: What is the Enterprise plan?
- Anthropic Enterprise billing guide
- Anthropic Claude Team plan guide
- Anthropic commercial data roles and model-training policy
- Anthropic US-only inference guide
The bottom line
Claude Enterprise is best understood as a governed, usage-metered AI work platform. The $20 seat gives an employee access; the organization then pays for the model work performed across chat, Claude Code, and Cowork.
That structure can be attractive for companies that want high-capacity work and tight controls, but it makes consumption forecasting essential. Claude Enterprise earns its place on a shortlist when its coding, long-form work, connectors, and governance fit your real workflows—and when the value of those workflows comfortably exceeds both the seat fee and the usage bill.
