Anthropic licensing covers the terms under which Anthropic makes its Claude models and applications available. For a software asset manager the portfolio falls into three groups. The first is Claude plans for people at work and at home: Team and Enterprise for organisations, Education for universities, and Pro and Max for individuals.[2] The second is the Claude API, billed per million tokens (MTok) by model.[3] The third is availability through cloud platforms: partner-operated on Amazon Bedrock and Google Cloud, and Anthropic-operated but marketplace-billed on Claude Platform on AWS and Microsoft Foundry.[3] Anthropic’s developer tools sit across these groups. Claude Code is proprietary software used under the Commercial Terms, and the client SDKs are released under the MIT License.[6][7]
Business and API use is governed by the Commercial Terms of Service, effective 2025-06-17. “Services under these Terms are not for consumer use”; Claude.ai and the individual plans fall under the Consumer Terms of Service.[1][8] The contracting party is Anthropic Ireland, Limited for customers in the European Economic Area, Switzerland or the UK, and Anthropic, PBC elsewhere.[1] None of these products uses per-core, per-processor or device metrics. Compliance work centres on seat counts, consumption, credits, use restrictions and data-use settings. This article summarises the position on 2026-10-01. Detailed prices and counting rules are in Anthropic Claude licensing.
Product families
| Family | What is licensed | How it is licensed | Guide |
|---|---|---|---|
| Claude Team | Claude, Claude Code and other features for a team | Standard or Premium seats, monthly or annual; “For teams of 2 to 150”[2] | Anthropic Claude licensing |
| Claude Enterprise | Organisation-wide Claude with administration and compliance features | Seat fee plus usage at API rates, billed annually; self-serve or sales-assisted[2] | Anthropic Claude licensing |
| Claude Education | University-wide plan for students, faculty and staff | Discounted rates, with dedicated API credits[2] | Anthropic Claude licensing |
| Claude Pro and Max | Individual subscriptions | Monthly or annual (Pro); from $100 per month (Max)[2] | Anthropic Claude licensing |
| Claude API | Model access for developers and applications | Per MTok of base input, cache writes, cache hits and output, by model[3] | Anthropic Claude licensing |
| Cloud platforms | Claude through Amazon Bedrock, Google Cloud, Claude Platform on AWS and Microsoft Foundry | Billed by the cloud provider, or by marketplace in Claude Consumption Units[3] | This article |
| Claude Code | Agentic coding tool | Proprietary; “Use is subject to Anthropic’s Commercial Terms of Service”; included in paid plans[6][2] | This article |
| Client SDKs | API client libraries, for example the Python SDK | MIT License[7] | This article |
Claude plans
The Team Standard seat costs $20 per seat per month billed annually, or $25 billed monthly. The Premium seat, with “5x more usage than standard seats”, costs $100 annually or $125 monthly.[2] Enterprise “Pricing is $20 per seat per month plus usage billed at API rates, so cost scales with the models and tasks your team runs.”[2] Enterprise seats give access but “doesn’t include any usage”. New Enterprise contracts use a single Enterprise seat, and legacy seat types move to it at renewal.[10] Pro costs $17 per month with an annual subscription ($200 billed up front), or $20 billed monthly. Max offers 5x or 20x more usage than Pro.[2] Max 5x costs $100 per month and Max 20x $200 per month.[16]
Claude API
As of 2026-10-01 the current models listed on the pricing page are Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5 and Claude Haiku 4.5. Base input and output prices per MTok are $10 and $50 for Fable 5.1, $4 and $20 for Opus 5.5, $2 and $10 for Sonnet 5.5, and $1 and $5 for Haiku 4.5.[3] The Batch API gives “a 50% discount on both input and output tokens”. Claude 4.6 and later models include the full 1M-token context window at standard pricing.[3]
Claude Code and SDKs
The Claude Code repository’s license file reads: ”© Anthropic PBC. All rights reserved. Use is subject to Anthropic’s Commercial Terms of Service.”[6] Claude Code is therefore not open source, even though its repository is public. The pricing page lists Claude Code as a feature of the Pro, Max, Team and Enterprise plans.[2] The Python SDK, by contrast, carries the MIT License text and a 2023 Anthropic, PBC copyright notice.[7] An inventory should record the two differently: Claude Code as a proprietary tool tied to a plan or API account, and the SDKs as open source components.
Metrics
The catalog records Anthropic’s metrics under Anthropic’s own names. The definitions are in the linked rows, and counting rules are in the guide.
| Metric | Where it applies | Role |
|---|---|---|
| Standard seat and Premium seat | Claude Team | Per member; Team plans have a minimum of two members[11] |
| Enterprise seat | Claude Enterprise | Access only; usage billed separately at API rates[10] |
| Usage credits | Paid Claude plans | Continued use at standard API rates after a limit is reached[2] |
| Base input tokens, cache writes, cache hits and refreshes and output tokens | Claude API; Enterprise usage | Price per MTok by model[3] |
| Claude Consumption Unit | Claude Platform on AWS; Claude in Microsoft Foundry | Marketplace invoicing unit[3] |
API rate limits are set at the organisation level, and organisations move between usage tiers automatically (API usage tiers and monthly spend caps).[13] As with other generative AI vendors, Anthropic publishes no conversion between seats and tokens.
Purchasing programs and agreements
- Published rates. API customers pay “at the rates specified on the Model Pricing Page, unless otherwise agreed by the parties”. Anthropic may update published rates, “to be effective the earlier of 30 days after the updates are posted by Anthropic or Customer otherwise receives Notice”.[1]
- Credits. Anthropic may require prepayment in credits, which are governed by the Supplemental Credit Terms.[1] Credits “are non-refundable”. “Usage Credits expire one calendar year from the date Anthropic sends the Confirmation Notice or otherwise issues the Credits”, and credits also expire if the account is closed.[4]
- Payment and billing. For the API, “Billing is based on actual monthly usage” and all payments are in USD. Credit cards are accepted for standard accounts, and Enterprise customers can arrange invoicing.[3] On the plan side, invoicing and net terms are listed only for sales-assisted Enterprise plans. Team is billed monthly or annually and Enterprise annually.[2]
- Discounts. “Volume discounts may be available for high-volume users. These are negotiated on a case-by-case basis.”[3] Sales-assisted Enterprise plans list “Tiered incentives on committed spend” and AWS Marketplace availability.[2]
- Reserved capacity. “Priority Tier capacity commitments are no longer available for purchase”; existing commitments run to their end date (program).[12]
The catalog’s program rows include the Claude Team plan, Claude Enterprise plan, Claude Education plan, Claude for Nonprofits and Claude Pro and Max.
Support and maintenance
Anthropic sells no separate maintenance product. Models and applications are updated as part of the service, and the Commercial Terms provide the Services “AS IS” and “AS AVAILABLE” except as the Terms state otherwise.[1] The pricing documentation lists “Dedicated support” among the options for enterprise customers.[3] Sales-assisted Enterprise plans list “Customer success support at certain spend thresholds”.[2]
For the API, the product lifecycle is the model lifecycle. Anthropic uses four states: Active, Legacy, Deprecated and Retired. Requests to a retired model fail. Anthropic notifies customers with active deployments, “providing at least 60 days’ notice before model retirement for publicly released models”.[5] The status table gives each active model a tentative retirement date. For claude-opus-5-5 it is “Not sooner than September 22, 2027”. These dates apply to Anthropic-operated platforms. “Partner-operated platforms (Amazon Bedrock and Google Cloud) set their own retirement schedules”.[5] Anthropic also states that it “has committed to long-term preservation of model weights”.[5] To find usage of deprecated models, Anthropic suggests exporting usage data from the Usage page of the Claude Console.[5]
Audit and compliance
The Commercial Terms have no customer licence audit clause of the on-premises kind. The nearest provision is a duty to cooperate: “Customer must cooperate with reasonable requests for information from Anthropic to support compliance with its Usage Policy, including to verify Customer’s identity and use of the Services.”[1] Compliance otherwise rests on these terms:
- Use restrictions. Customers may not use the Services “to build a competing product or service, including to train competing AI models or resell the Services except as expressly approved by Anthropic”, or reverse engineer or duplicate them (Commercial use restrictions).[1] The Usage Policy, effective 2025-09-15, applies to all users and allows Anthropic to throttle, suspend or terminate access for violations.[9]
- Account responsibility. “Customer is responsible for all activity under its account.”[1] The Consumer Terms forbid sharing account login information or API keys.[8]
- Suspension. Anthropic may suspend access for risks to the Services, for violations of the compliance, policy or use-restriction sections, or where the law prohibits providing them. “Failure to pay Anthropic all amounts owed when due may result in suspension or termination of Customer’s access to the Services.”[1]
- Termination. “Each party may terminate these Terms at any time for convenience with Notice, except Anthropic must provide 30 days prior Notice.” Either party may terminate for uncured material breach on 30 days’ notice.[1]
- Changes. Anthropic may update the Terms with effect 30 days after posting or Notice. Updates required by changes in law take effect immediately, and “Changes will not apply retroactively.”[1]
Indemnity and output
The customer “(a) retains all rights to its Inputs, and (b) owns its Outputs”, and “Anthropic may not train models on Customer Content from Services” (No training on Customer Content).[1] Anthropic defends customers against third-party claims that the customer’s “paid use of the Services (which includes data Anthropic has used to train a model that is part of the Services)” or authorised Outputs violate intellectual property rights. The indemnity excludes claims arising from the customer’s own modifications, breach of the Agreement and similar causes. The limitations of liability “do not apply to either party’s obligations under Section K (Indemnification)”.[1] The indemnity covers paid use only, so free or promotional use falls outside it.
Cloud, hosting and territorial rules
Claude is a hosted service, so there is no bring-your-own-license question. What matters is which entity operates and invoices each route:
| Route | Operated by | Invoiced by | Notes |
|---|---|---|---|
| Claude API and Claude plans | Anthropic | Anthropic | Global by default; US-only inference costs 1.1x for Claude 4.6 and later models[3] |
| Claude in Amazon Bedrock | AWS | AWS | Runs “on AWS-managed infrastructure with zero operator access”[14] |
| Claude on Google Cloud | Google Cloud | Google Cloud | Regional and multi-region endpoints carry a 10% premium over global endpoints[3] |
| Claude Platform on AWS and Claude in Microsoft Foundry | Anthropic | AWS Marketplace or Azure Marketplace, in Claude Consumption Units | US Data Zone deployments on Azure carry the same 1.1x multiplier as US-only inference[3] |
Use of the Services must follow Anthropic’s Supported Regions Policy, which the Commercial Terms incorporate.[1] Customers may not export or provide access to the Services where U.S. or other applicable law prohibits it.[1] For API data, Anthropic’s privacy documentation states that inputs and outputs are deleted from its backend within 30 days, subject to exceptions such as zero data retention arrangements, Usage Policy enforcement and legal requirements.[15] On the individual plans, the comparison table lists model training as “Opt-out”.[2]
Recent licensing changes
| Date | Change |
|---|---|
| 2024-03-04 | Supplemental Credit Terms took effect[4] |
| 2025-06-17 | Current Commercial Terms of Service took effect[1] |
| 2025-09-15 | Usage Policy (formerly Acceptable Use Policy) took effect[9][1] |
| 2025-10-08 | Consumer Terms of Service took effect (EEA and Switzerland version)[8] |
| 2026-06-09 | claude-mythos-preview deprecated; retirement date to be announced[5] |
Undated changes recorded in the help and API documentation include a single Enterprise seat for new Enterprise contracts, with legacy seat types converting at renewal, and the withdrawal of Priority Tier capacity commitments from sale.[10][12]
Published API prices can change on 30 days’ notice, so they should be recorded with their retrieval date.[1]
Guides
| Article | What it covers |
|---|---|
| Anthropic Claude licensing | Team, Enterprise, Education, Pro and Max plans; seat counting; Enterprise usage billing; API token prices, prompt caching, batch and data residency; rate limits; data use and retention; usage restrictions; cloud and marketplace billing; programs |
| Anthropic Commercial Terms and Usage Policy | Commercial versus consumer terms, the documents the Commercial Terms incorporate, use restrictions, Usage Policy layers and High-Risk Use Case Requirements, Service Specific Terms (Claude for Work, beta, fine-tuning, marketplaces, Development Partner, Covered Models), Supported Regions Policy and the DPA |
Related articles: OpenAI licensing for a comparable vendor, generative AI licensing for the cross-vendor concepts and subscription and consumption licensing for the general model.
Out of scope
This article does not cover Claude Managed Agents session-runtime pricing, Fast mode rates, tool prices, models offered by invitation only, Claude for Government, the Data Processing Addendum and Service Specific Terms in detail, or negotiated Enterprise and volume prices, which Anthropic does not publish. Pricing on Amazon Bedrock and Google Cloud is set on the cloud providers’ own pages.[3]
Catalog rows
- Metrics: Standard seat, Premium seat, Enterprise seat, Usage credits, Base input tokens, Output tokens, Claude Consumption Unit
- Programs: Claude Team plan, Claude Enterprise plan, Claude Platform on AWS and Claude in Microsoft Foundry, Priority Tier
- Rules: Team plan two-member minimum and 150-seat maximum, Team plan billed on members at cycle start, Enterprise seat fee plus usage at API rates, No training on Customer Content, Commercial use restrictions, API usage tiers and monthly spend caps, Marketplace usage invoiced in CCU