Claude
Claude is Anthropic's general-purpose AI assistant for chat, writing, file analysis, coding and delegated agent work. It runs on the web, mobile, desktop, the terminal and an API, with a permanent free plan.
What is Claude?
Claude is the AI assistant built by Anthropic PBC, a company that presents itself first as an AI safety and research lab and only then as a product vendor. The assistant is trained with Constitutional AI, which Anthropic describes as a method for making it safe, accurate and secure, and the homepage frames it as a thinking partner rather than an answer machine.
Underneath sit four models you choose between according to the job: Fable 5 for long-running agentic work, Opus 5 for complex analysis and deep research, Sonnet 5 for everyday tasks, and Haiku 4.5 when speed and cost matter most. Around them Anthropic has built a family of products that share one subscription: Claude for chat, Claude Code for agentic coding in the terminal and in IDEs, Claude Cowork for delegated and scheduled tasks, plus Claude Design, Claude Science and a beta Claude Security. Extensions reach into Chrome, Slack and Microsoft 365.
In practice the assistant reads what you give it - PDFs, Word files, spreadsheets, images, CSVs - searches the web when asked, and returns work rather than paragraphs: interactive artifacts, dashboards, diagrams and research reports carrying citations in APA, MLA or Chicago style. Projects keep context per topic, memory carries across conversations, and you can import your memory from another AI provider. Connectors and remote MCP plug Claude into third-party tools, with a marketplace, plugins and Skills extending the same idea.
Developers get a separate route through platform.claude.com, where the models are billed by the token and composable APIs support agents at scale. The same models are resold through AWS Bedrock, Google Cloud Vertex AI and Microsoft Foundry, which is how regional data residency is offered across the United States, Europe, Canada and Asia-Pacific.
Anthropic points to Norges Bank Investment Management, Novo Nordisk and the European Parliament as reference customers. The counterweight is usage: every plan, paid ones included, runs against a rolling five-hour window with weekly caps on top, and there is no published message count.
What it does
- Hold a working conversation by text or voice, with extended thinking for harder problems
- Write, edit and proofread long-form content, and keep it organised in per-topic projects
- Generate, debug, explain and review code, and act directly in the terminal or IDE through Claude Code
- Read uploaded PDFs, Word files, spreadsheets, images and CSVs, then compute and visualise what is in them
- Build interactive artifacts - documents, dashboards, diagrams - that can be shared with people who have no Claude account
- Search the web and return research reports with proper citations in APA, MLA or Chicago style
- Run scheduled, delegated tasks in Claude Cowork and hand back a finished deck, document or spreadsheet
When to use Claude / When not to
A quick filter to help you decide if Claude is the right fit.
When to use Claude
- Software engineers and DevOps teams, who get agentic coding in the terminal and IDE extensions for VS Code, Cursor and JetBrains at no extra cost on any paid plan
- Analysts and researchers who work from documents and spreadsheets and need sourced, citable reports rather than a plain chat answer
- Writers, editors and content teams handling long-form drafting, rewriting and proofreading across several formats
- Regulated organisations in finance, healthcare, legal and government, drawn by SOC 2 Type 2, ISO/IEC 27001 and regional data residency
- Universities, teachers and students, covered by a dedicated Education plan spanning students, faculty and staff
When not to use Claude
- Anyone looking for image, video or music generation: none is advertised on the pages reviewed
- Under-18s, who are excluded by the consumer terms
- Users in China, Russia, Iran, Hong Kong, Macau and the other territories missing from the published list of 184 supported countries
- Teams that need strictly predictable bills, since Enterprise adds usage at API rates on top of the seat price
- Organisations that require a self-hosted, on-premise deployment: the only options are Anthropic's own service or the three major clouds
How to use Claude
A typical end-to-end flow, from setup to results.
- Go to claude.ai and create an account with Google, an email address or SSO
- Start on the Free plan at $0 - no payment and no time limit - to see whether the assistant fits your work
- Pick your surface: browser, macOS or Windows desktop app (Linux documented), iOS, Android, or ChromeOS through Google Play
- Ask a question, or drop in the files you want read - PDF, Word, Excel, images, CSV
- Choose the model for the task: Haiku for speed, Sonnet for everyday work, Opus for complexity, Fable for long-running jobs
- Turn on extended thinking, web search or voice mode when the task calls for it
- Group related work in a Project so context persists instead of being re-explained
- Connect your tools through connectors, remote MCP, Slack, Google Workspace or Microsoft 365
- For engineering, install Claude Code - included on every paid plan - or the VS Code and JetBrains extensions
- Open your account settings to opt out of model training, and switch to platform.claude.com if you would rather pay per token
Pros & Cons
Pros
- A genuinely usable free plan: web, iOS, Android and desktop chat, web search, memory across conversations, file creation and remote MCP, at no cost
- One subscription covers chat, agentic coding and delegated tasks - Claude Code is included on every paid plan
- A tiered model range lets you trade cost against capability instead of paying top rates for every request
- A substantial compliance file: SOC 2 Type 2, ISO/IEC 27001, 27017 and 27018, CSA STAR, a public DPA and a HIPAA-ready offering
- Regional data residency across the United States, Europe, Canada and Asia-Pacific, available through the three major clouds
- Output ownership is assigned to the user, and commercial terms bar training on Customer Content outright
- Published API pricing per token, with 50% off batch processing and prompt caching to cut real costs
Cons
- Usage limits apply on every plan including paid ones, on a rolling five-hour window with weekly caps, and no message count is published
- On consumer accounts, conversations feed model training by default - the opt-out is yours to find and switch on
- Even after opting out, conversations flagged for safety review and anything you report stay in use
- Enterprise bills $20 per seat plus usage at API rates, which makes the final invoice hard to forecast
- Enterprise is sold on an annual subscription only
- Access is restricted to a published list of 184 countries and regions, leaving out China, Russia, Iran, Hong Kong and Macau among others
- Payments are generally non-refundable outside the cases set by the terms or by local law
Pricing & Plans
A permanent free plan is available at $0, described as free for everyone. The cheapest paid entry point is the Pro plan at $17 per month on an annual commitment, billed as $200 up front, rising to $20 per month if paid monthly. Prices exclude applicable tax and are subject to change at Anthropic's discretion.
- $0 - chat on web
- iOS
- Android and desktop
- code generation
- data visualisation
- web search
- memory across conversations
- file creation and code execution
- $17 per month billed annually ($200 up front) or $20 per month - everything in Free plus more usage
- Claude Code
- Claude Cowork
- Claude Design
- Claude Science
- unlimited projects
- Research
- more models and Claude for Microsoft 365
- from $100 per month - everything in Pro plus 5x or 20x the usage of Pro per session
- higher output limits
- early access to new features and priority at peak times
- $20 per seat per month billed annually ($25 monthly) for a Standard seat
- $100 per seat per month annually ($125 monthly) for a Premium seat
- for teams of 2 to 150 - enterprise search
- central billing
- SSO
- connector admin controls
- managed desktop deployment and no model training on your content by default
- $20 per seat per month plus usage at API rates
- annual only - spend limits
- fine-grained role-based access
- SCIM
- audit logs
- a Compliance API
- custom retention
- network access control
- a university-wide plan covering students
- faculty and staff at discounted rates
- with dedicated API credits and a learning mode
- Fable 5 at $10/$50 per million input/output tokens
- Opus 5 at $5/$25
- Sonnet 5 at $2/$10 (introductory until 31 August 2026
- then $3/$15) and Haiku 4.5 at $1/$5
- with batch processing at half price
Data, GDPR & hosting
A consolidated view of how Claude handles your data.
GDPR overview
GDPR compliance is claimed in concrete terms. The regional compliance page lists "GDPR (General Data Protection Regulation) with regional processing within EU/EEA" alongside a published Data Processing Addendum. Anthropic Ireland, Limited is named as the controller for the European Region, Anthropic PBC elsewhere. The privacy policy details the full set of rights - access, portability, erasure, rectification, objection, restriction and withdrawal of consent - with a one-month response window under EU and UK GDPR, extendable by two months. Transfers outside the EEA rely on adequacy decisions under Article 45 and standard contractual clauses under Article 46. Anthropic states it does not sell personal data, honours Global Privacy Control signals and makes no solely automated decisions with legal effect. Contacts are privacy@anthropic.com and dpo@anthropic.com. No Article 27 representative is named.
Who owns the data?
You keep your Inputs and Anthropic assigns you whatever rights it holds in the Outputs, so ownership sits with the user rather than the vendor. Commercial customers get the same deal in stronger terms: they retain all rights to their Inputs, own their Outputs, and Anthropic is contractually barred from training models on Customer Content. Consumer accounts are treated differently. There, Inputs and Outputs may be used to train and improve Anthropic's models unless you switch training off in your account settings, and two exceptions survive that opt-out: conversations flagged for safety review, and anything you report yourself through the feedback controls. A thumbs-up or thumbs-down stores the whole conversation as Feedback.
Reuse rights
Yes. Because Anthropic assigns its rights in the Outputs to you, you can reuse what Claude produces without asking permission first, including commercially. Two conditions attach. The assignment depends on your continued compliance with the terms and with the Usage Policy, which rules out a defined set of uses such as weapons development, CSAM and circumventing regulatory controls. And Anthropic makes no warranty about factual accuracy: outputs must be checked before they are relied on. The company also notes that similar or identical outputs may be generated for other users, so nothing you receive is exclusive to you. Commercial customers are explicitly stated to own their Outputs.
Data retention & training
Hosting summary
Personal data is transferred to Anthropic's servers in the United States, or to other countries outside the EEA and the UK, under adequacy decisions (Article 45 GDPR) and standard contractual clauses (Article 46). Anthropic Ireland, Limited is the controller for the European Region; Anthropic PBC elsewhere. Regional data residency is offered across Europe, the United States, Canada and Asia-Pacific, the latter including Japan, South Korea, Singapore, India and Australia. Regional endpoints guarantee that both storage and model processing stay inside the chosen region, while global endpoints route dynamically for availability at no price premium. Country-level options in Europe run through AWS Bedrock, Google Cloud Vertex AI and Microsoft Foundry. A subprocessor list is published. The claude.com domain resolves to a US IP address on Anthropic's own network, though that is an anycast edge node and not a storage location.
Where Claude works
Country-level availability.
Available in
Not available in
Things to keep in mind
Risks and trade-offs to weigh before adopting Claude.
- Cognitive dependence: a daily thinking partner can quietly replace the effort of working a problem through yourself
- Accuracy: Anthropic states it cannot guarantee the factual accuracy of outputs, and that inaccurate personal data inside a model may not always be correctable
- Privacy by default on consumer accounts: training is on unless you turn it off, and flagged or reported content stays in use even after you do
- Agentic permissions: depending on what you grant, outputs can send messages, modify files or act on third-party services on your behalf
- Third-party connectors: Anthropic does not control what connected services do with your data, and you are responsible for having the right to grant access
- Cost drift: API usage and Enterprise seats billed at API rates can grow well beyond a forecast
- Age verification may involve an identity document and facial geometry, which counts as biometric data in some jurisdictions
Setup & Integrations
Technical difficulty
Low for individuals. Creating an account on claude.ai with Google, email or SSO takes a minute and installs nothing; the desktop apps for macOS and Windows are direct downloads, and the mobile apps come from the usual stores. It rises with ambition: connectors and remote MCP need configuration and, on Team and Enterprise, administrator approval. Claude Code assumes comfort with a terminal or an IDE extension. API use requires a Console account and developer work, and enterprise rollout adds SSO, SCIM, IP allowlisting and managed desktop deployment.
Deployment
Apps stores
Integrations
Supported languages
Behind Claude
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What is Claude and who builds it?
Is there a free version?
How much does the cheapest paid plan cost?
Is Claude Code included in my subscription?
Will my conversations be used to train the models?
Who owns what Claude produces?
Is there a minimum age?
Which devices and platforms does Claude run on?
Is Claude available everywhere?
What compliance guarantees does Anthropic publish?
Should you pick Claude?
Claude is one of the few AI assistants where the company behind it is as much of the argument as the product. Anthropic sells safety research first and software second, and that shows in what is published: SOC 2 Type 2, ISO/IEC 27001, 27017 and 27018, CSA STAR, a public DPA, a HIPAA-ready offering and regional data residency across four regions. For a regulated buyer, that file is unusually complete.
The commercial design is equally deliberate. One paid subscription covers the chat assistant, agentic coding in the terminal and delegated background work, so teams are not buying three tools. The tiered model range means you can spend on capability only where a task needs it, and the API is billed transparently per token with batch processing at half price.
Two reservations deserve attention before committing. The first is usage: every plan, paid ones included, runs against a rolling five-hour window with weekly caps on top, and Anthropic publishes no message count, so how far a subscription goes is genuinely hard to predict. The second is privacy on consumer accounts, where conversations feed model training by default and the opt-out is something you must go and switch on yourself - and even then, flagged and reported content stays in use. Commercial customers get the opposite regime, with training on Customer Content contractually barred, which makes the plan you choose a privacy decision rather than only a budget one.
Add the restriction to 184 supported countries and the Enterprise bill that scales with API usage, and the picture is clear: a capable, well-documented assistant whose fine print rewards reading.
- Choosing a selection results in a full page refresh.
- Opens in a new window.