Claude Enterprise Explained: What MNCs Get That the Pro Plan Doesn't

Most people who look at Claude's pricing page make the same mistake.

They see Pro at $20/month and Enterprise at "contact sales" and assume Enterprise is Pro with a few extra features bolted on. A bigger usage limit. Maybe some admin controls. Probably not worth the price jump for most companies.

That assumption is wrong and it is the wrong frame entirely.

Claude Pro is built for individual productivity. You use Claude. Claude answers. The session ends. Nobody else in your organisation knows it happened. Nothing is logged. Nothing is governed. Nothing is audited.

Claude Enterprise is built for organisational governance. Not what Claude can do what your organisation can control, monitor, audit, and prove to a regulator. Those are fundamentally different problems. And for any MNC operating in a regulated industry, the difference is not a feature gap. It is the gap between a tool you can deploy and a tool your security team will never approve.

This guide covers every meaningful difference between Pro and Enterprise what the features actually do, why they matter in a real organisational context, and how to determine whether your company needs Enterprise or whether Team is sufficient.

The Pricing Structure First Because It Matters

Before the features, the cost architecture deserves clarity because it is genuinely different from what most people expect.

Claude Pro: $20/month per user. Fixed price. Usage is included within the plan's limits. Hit the limit, wait or upgrade.

Claude Team: $25/seat/month (annual) or $30/month (monthly). Minimum 5 seats. Usage is included within per-seat limits. Optional Premium seats at $100/seat/month add Claude Code and Cowork with 6.25x more usage.

Claude Enterprise (Self-Serve): Starts at approximately $20/seat/month billed annually, minimum 20 seats. Usage is billed separately at standard API rates on top of the seat fee. This is a fundamentally different billing model the seat fee covers access and governance features, and every token your team uses is an additional line item.

Claude Enterprise (Sales-Assisted): Custom pricing, minimum 50 seats, tailored contract terms, invoicing, usage commitments, and the HIPAA-ready offering with a signed Business Associate Agreement. This is the tier that Fortune 500 legal, compliance, and procurement teams actually engage with.

The implication of the usage-at-API-rates model: Enterprise costs scale with actual use rather than being capped at a plan limit. For high-volume teams, this is often cheaper than the per-seat limit model. For teams that need governance features but have moderate usage, the economics need to be modelled specifically. There is no universal answer about whether Enterprise is more or less expensive than Pro it depends entirely on volume and team size.

Context Window: 1 Million Tokens vs 200K

The context window is the clearest functional difference, and the one most business stakeholders underestimate.

Both Claude Opus 4.7 and Claude Sonnet 4.6 run with a 1 million token context window on Enterprise compared to 200K on Pro and Team. Claude Haiku 4.5 runs at 200K across all tiers.

One million tokens is approximately 750,000 words. In practical terms:

A Pro user can load one large document a 200-page contract, a year of financial statements, a substantial research report and ask Claude to reason across it.

An Enterprise user can load an entire knowledge base simultaneously. Hundreds of sales transcripts. Dozens of 100-page documents. 100,000 lines of code. A full legal case file including exhibits, correspondence, and precedent documents. A compliance review covering every policy document the organisation has produced in the past three years.

The difference is not about convenience. It is about the class of problems the tool can solve. Many of the highest-value enterprise AI use cases M&A due diligence, regulatory compliance review, large codebase analysis, cross-document research synthesis are technically impossible at 200K tokens. They are routine at 1 million.

SSO and Domain Capture The Identity Foundation

Single Sign-On (SSO) and domain capture allow organisations to centralise login and auto-provision users with the company domain.

On Pro, every user authenticates individually with their own credentials. There is no organisational control over who has access, what account they're using, or whether a departing employee still has active access.

On Enterprise, SSO enforces that every user authenticates through the company's identity provider Okta, Microsoft Entra, Google Workspace, or equivalent. Domain capture means any user signing up with a company email address is automatically routed to the organisational account, preventing the shadow AI problem where individuals create personal Claude accounts using their work email.

The security implication: when an employee leaves, their Claude access terminates through the same offboarding process that terminates their email, their Slack, and their CRM access. No separate Claude deprovisioning step. No window where a departing employee still has access to organisational Claude sessions.

For a 5,000-person organisation processing any volume of sensitive information through Claude, the ability to guarantee that access control is real-time and complete is not a nice-to-have. It is the minimum requirement for any security team sign-off.

SCIM Automating the User Lifecycle at Scale

SCIM (System for Cross-domain Identity Management) automates user provisioning and de-provisioning across the identity provider.

Without SCIM, adding 500 new employees to Claude Enterprise requires manual seat creation for each. Removing access requires manual action per departing user. At MNC scale where dozens or hundreds of employees join or leave every month manual provisioning is not operationally viable.

With SCIM, user lifecycle management is automated entirely. New hires are provisioned to Claude the moment their identity provider account is created. Departing employees are deprovisioned instantly. Role changes a promotion, a department transfer, a project assignment update Claude access automatically based on the identity provider group membership.

The compliance implication is significant. Regulators auditing access controls want evidence that access is current, accurate, and promptly revoked. Manual provisioning processes have inherent gaps. SCIM provides the automated, auditable lifecycle that compliance teams require.

Victorian railway signal box editorial cartoon showing Claude Pro's uncontrolled user access versus Claude Enterprise's SCIM-automated provisioning and deprovisioning at MNC scale

Audit Logs The Feature That Makes Deployment Legal in Regulated Industries

Audit logs provide a full record of user actions, system events, and data access. Audit logging under the Enterprise plan captures user authentication events, model calls with associated metadata, and file interactions.

On Pro, there is no audit trail. A user has a conversation with Claude, shares sensitive data, receives an output and there is no record of what happened that an administrator can access, export, or present to a regulator.

On Enterprise, every significant action generates an auditable record. Who accessed Claude. When. What models were called. What files were interacted with. The metadata is available to administrators through the Compliance API for programmatic export.

For financial services firms subject to MiFID II or SOX, healthcare organisations under HIPAA, legal firms with client confidentiality obligations, government contractors with FedRAMP requirements audit logs are not optional. They are the mechanism by which the organisation demonstrates to regulators that AI use was governed, monitored, and controlled. Without them, enterprise AI deployment in regulated industries is not a risk management question. It is a non-starter.

The Compliance API and Analytics API

The Compliance API provides programmatic access to activity logs, chat histories, and file content by user and time range. The Analytics API provides access to aggregated adoption metrics across the organisation.

These two APIs transform Claude from a tool your employees use into an asset your organisation governs.

The Compliance API means a legal or compliance team can programmatically pull every Claude interaction involving a specific user, during a specific time range, involving specific file types without manual extraction from an admin dashboard. For a regulatory investigation, an internal audit, or a legal discovery request, this is the difference between a response that takes weeks and one that takes hours.

The Analytics API means a CIO or AI governance committee can track actual adoption patterns across the organisation which departments use Claude most, which use cases are growing, where usage is concentrated — without relying on self-reported data. That visibility is what makes responsible scaling possible. You cannot govern what you cannot measure.

Role-Based Access Controls Precision Governance at Scale

Role-based access controls allow granular permission assignment across the organisation. Enterprise RBAC controls six specific capabilities SSO enforcement, SCIM provisioning, custom RBAC roles, Chrome settings, tenant restrictions, and group-based access policies.

On Pro, every user has the same access. There is no mechanism for a Chief Risk Officer to have different Claude capabilities than an intern. No way to restrict certain departments from using specific models. No way to ensure that the team handling regulated client data cannot also export that data through Claude in ways the organisation hasn't approved.

On Enterprise, access is configured by role. Portfolio analysts access client data within their designated business units. Senior managers have cross-portfolio visibility. Compliance officers have read access across all interactions. Specific teams can be restricted from using specific connectors or MCP integrations. High-risk capabilities are gated behind approvals that the standard user never sees.

The financial services example from AWS's own Claude Enterprise documentation illustrates it precisely: a financial services organisation uses role-based permissions so that portfolio analysts only see client performance data within their business units, senior managers see across portfolios, and compliance officers maintain audit access across all interactions with automated trails tracking every request involving regulated data.

Custom Data Retention and Zero Data Retention

Without Zero Data Retention, Anthropic retains interaction data for 30 days by default under standard operational terms. Enterprise customers can configure custom data retention how long data is stored and when it is deleted and qualify for Zero Data Retention (ZDR).

With ZDR, no conversation data is written to disk after the session. Abuse checks still run in-pipeline during the session, but nothing persists after it ends. The session never existed from a storage perspective.

This is the configuration that makes Claude viable for the most sensitive enterprise workloads. A team processing personally identifiable health information. A legal team working through client-privileged documents. A financial institution running analysis on non-public market data. In each case, the fundamental requirement is that the AI does not retain copies of what it processed. ZDR is the mechanism that satisfies that requirement.

HIPAA-Ready Configuration and the BAA

The sales-assisted Enterprise tier includes HIPAA-ready configurations with a signed Business Associate Agreement (BAA) available for qualifying customers.

For any organisation handling Protected Health Information under HIPAA, a BAA is legally required before using any third-party service that processes that data. The BAA establishes Anthropic as a Business Associate under HIPAA, commits them to specific data handling and security obligations, and makes them jointly responsible for HIPAA compliance in relation to the data they process.

Claude Pro has no BAA. Claude Team has no BAA. HIPAA-eligible organisations cannot legally deploy Pro or Team for any workflow involving PHI regardless of how secure the actual data handling is in practice. The BAA is the legal instrument, and it is only available through the sales-assisted Enterprise tier.

Healthcare organisations, health insurance companies, pharmaceutical firms, and any other entity subject to HIPAA that wants to use Claude for clinical, administrative, or research workflows involving PHI have one option: Enterprise, sales-assisted, with a signed BAA.

Courtroom editorial cartoon showing healthcare executive with Claude Pro facing HIPAA compliance hearing without a BAA versus Enterprise executive presenting signed BAA and audit log documentation

Claude Security The April 2026 Enterprise Exclusive

Claude Security launched in public beta on April 30, 2026, exclusively for Enterprise users. Using Claude Opus 4.6, Anthropic's own team found over 500 vulnerabilities in production open-source codebases in internal testing, including bugs that had gone undetected for years.

Traditional code security tools SAST, DAST, SCA match code against known vulnerability patterns. Claude Security does something different: it analyses code behaviour, traces data flows, and identifies complex vulnerabilities that pattern-matching misses. It runs each finding through a multi-stage verification process, assigns severity and confidence ratings, and requires human approval before any fix is applied to production.

The announcement caused immediate market reaction. CrowdStrike, Zscaler, and Datadog each fell around 11% on the first full trading day following the Claude Security release, as investors priced in potential disruption to the SAST market.

For MNCs with large engineering teams, this is a meaningful capability arriving as an Enterprise exclusive. The practical implication: a 500-engineer organisation running Claude Security across its codebase gets vulnerability analysis that conventional tools would either miss entirely or bury in false positives. The multi-stage verification process directly addresses the false positives problem that makes AI-driven security analysis operationally impractical in most environments.

Team and Max access is coming but as of the May 2026 launch, this is an Enterprise-only capability.

Infrastructure Deployment: AWS Bedrock, Google Vertex, Azure

Enterprise customers can deploy Claude on their own cloud infrastructure AWS Bedrock, Google Cloud Vertex AI, and Microsoft Azure AI Foundry rather than Anthropic's cloud.

This matters for three categories of organisations:

Data sovereignty requirements. Some regulated organisations particularly in financial services and government cannot route data through a third-party AI provider's infrastructure regardless of the security controls. They must process data within infrastructure they control. Bedrock, Vertex, and Azure deployments satisfy this requirement by keeping Claude on the customer's own cloud.

Existing cloud commitments. Large organisations typically have multi-million dollar cloud commitments with AWS, Google, or Microsoft. Deploying Claude through their existing cloud provider applies usage against committed spend, reducing effective cost and simplifying procurement.

IT governance. Organisations with mature cloud governance frameworks VPC configurations, private endpoints, DLP policies, data loss prevention at the network layer can apply those frameworks to Claude usage when it runs on their own infrastructure. That governance is not available when using Anthropic's hosted service.

What Enterprise Still Does Not Cover

Honest disclosure matters, because two significant gaps affect MNC deployment decisions.

Cowork activity is explicitly excluded from Audit Logs, Compliance API, and Data Exports as of May 2026. This applies across every plan tier, including Enterprise. Organisations running Cowork for automated workflows cannot pull compliance reports showing what files a Cowork session accessed or what actions it took through Anthropic's native tools.

The practical implication: regulated organisations using Cowork for workloads subject to SOX, HIPAA, or PCI compliance need to implement supplementary observability tooling OpenTelemetry or similar to achieve the audit coverage that Anthropic's native compliance mechanisms don't yet provide for Cowork specifically. This is a known gap, and the security community is actively monitoring whether Anthropic will close it.

The second gap: RBAC controls do not cover Chrome, plugins, MCP servers, or connectors. Role-based access controls six capabilities but the MCP server and connector layer that gives Claude access to external systems sits outside the RBAC framework. Organisations need supplementary controls managed allowlists, MDM policy enforcement, network-layer restrictions to govern which external tools Claude can reach on behalf of employees.

Neither gap prevents Enterprise deployment. Both gaps require security teams to build compensating controls rather than relying exclusively on Anthropic's native governance.

Pro vs Team vs Enterprise: The Decision Framework

Choose Pro if you are an individual or very small team with no compliance requirements, no need for shared governance, and primarily individual productivity use cases. Pro is the right tool for individual power users. It is not an organisational deployment.

Choose Team if you have a team of 5–150 people, operate in a non-regulated industry, do not require SSO or SAML, and want collaborative Projects, shared context, and organisational billing without enterprise procurement complexity. Team is where most growing businesses should start.

Choose Enterprise (Self-Serve) if you have 20+ users, need SSO, SCIM, audit logs, and compliance API, operate in a regulated industry, or require custom data retention. The self-serve tier provides the governance layer without the sales cycle.

Choose Enterprise (Sales-Assisted) if you have 50+ users, need HIPAA-ready configuration with a BAA, require tailored contract terms or invoicing, want usage commitments, or need the full compliance and security stack for a regulated industry deployment at scale.

The signal that Enterprise is required rather than optional: if your legal, compliance, or security team is involved in the decision. Pro and Team are products you deploy. Enterprise is a product you procure. The procurement process exists because the governance requirements of regulated industries demand it.

FAQ

What is Claude Enterprise and how is it different from Claude Pro? Claude Enterprise is Anthropic's organisational tier for large companies with compliance, security, and governance requirements. The difference from Pro is not additional usage it's a different product architecture. Enterprise adds SSO, SCIM, audit logs, Compliance API, custom data retention, Zero Data Retention, HIPAA-ready configurations, role-based access controls, and a 1 million token context window. Pro has none of these. Pro is for individual productivity. Enterprise is for organisational governance.

How much does Claude Enterprise cost? The self-serve Enterprise tier starts at approximately $20/seat/month billed annually, with a minimum of 20 seats. User reports put pricing at approximately $60/seat/month for standard deployments, though this varies by contract. Usage is billed separately at standard API rates on top of the seat fee. Sales-assisted Enterprise has custom pricing with a minimum of 50 seats. The total cost depends on seat count, usage volume, and negotiated terms.

Does Claude Enterprise include HIPAA compliance? HIPAA-ready configuration and a signed Business Associate Agreement (BAA) are available through the sales-assisted Enterprise tier only. The BAA makes Anthropic a Business Associate under HIPAA and is legally required before processing Protected Health Information. Neither Pro nor Team includes a BAA. Organisations handling PHI cannot legally use Pro or Team for those workloads regardless of how secure the actual data handling is in practice.

What is Zero Data Retention and who needs it? Zero Data Retention (ZDR) means no conversation data is written to disk after a session ends. Without ZDR, Anthropic retains interaction data for 30 days by default. ZDR is available as an Enterprise addendum and is the correct configuration for any team processing PHI, financial data, client-privileged information, or any regulated data category. It is one of the most common requirements from security and legal teams evaluating Claude for sensitive enterprise workflows.

What is Claude Security and which plans include it? Claude Security is Anthropic's code vulnerability scanner powered by Claude Opus 4.6, launched in public beta on April 30, 2026. It analyses code behaviour and data flows to identify complex vulnerabilities that traditional SAST tools miss, running a multi-stage verification process before surfacing findings. As of launch it is available exclusively on Enterprise. Access for Team and Max is planned but not yet available.

Can Claude Enterprise be deployed on AWS, Google Cloud, or Azure? Yes. Enterprise customers can deploy Claude on AWS Bedrock, Google Cloud Vertex AI, and Microsoft Azure AI Foundry. This allows organisations with data sovereignty requirements, existing cloud commitments, or mature cloud governance frameworks to run Claude within infrastructure they control rather than Anthropic's hosted service.

What are the current limitations of Claude Enterprise? Two significant gaps affect MNC deployment: Cowork activity is excluded from Audit Logs, Compliance API, and Data Exports as of May 2026, meaning regulated organisations using Cowork for compliance-sensitive workflows need supplementary observability tooling. Additionally, RBAC controls do not cover MCP servers, connectors, or Chrome these require separate governance controls at the network or MDM layer.

How many seats are required for Claude Enterprise? The self-serve Enterprise tier requires a minimum of 20 seats with an annual commitment. The sales-assisted tier requires a minimum of 50 seats. Teams with 5–150 members that do not require Enterprise-grade compliance features can use the Team plan, which starts at 5 seats.