Verdict: a strong everyday assistant for anyone who reads and writes long documents; start on Free and check the Model Improvement privacy setting before you paste work material. Move to Pro ($20 a month) when the limits stop you several times a week, and to Team once colleagues use it for work.
Other good fits
Getting finished files back (Word, Excel, slides) instead of text to copy out.
People who want one subscription that also covers a coding agent and bigger delegated tasks.
+ Pros
Read pros
Claude is Anthropic's assistant on the web, on desktop and on your phone. Hand it a long PDF, a messy spreadsheet or a rough draft, and you get a careful read, a rewrite or a working file back. Since 16 September 2026 Anthropic has been merging chat and Cowork into one Claude, so bigger multi-step jobs (a report plus matching slides, say) start from the same message box. Pro and Max get this first.
Pick it when…
You work with long documents and want a reader that keeps track of detail across many pages.
You want files back, not just text: Claude creates and edits documents, spreadsheets and slides, and runs code to do it.
You want one subscription that also covers a coding agent (Claude Code) and bigger hand-over tasks (Cowork).
− Cons
Read cons
Claude still gets facts, numbers and references wrong, so check anything that leaves your desk. Free has tight limits, no Opus and no Research, and the merged chat-and-Cowork experience reaches Free and Team later than Pro and Max. On Free, Pro and Max, your chats and coding sessions train Claude only if Model Improvement is on. You choose at sign-up, so look before you paste anything sensitive, or use an incognito chat.
Pick something else when…
Your organisation runs Google Workspace and has approved Gemini.
You want the assistant most colleagues already know, with a cheaper paid tier: ChatGPT.
Answers must start from web sources with citations: Perplexity.
Data mustn't leave your own servers: Open WebUI with a local model.
Avoid if
Answers must lead with web sources: Perplexity (tool page) is built for that.
Data must stay on your own servers: run Open WebUI (tool page) with a local model.
Workflow
Describe a limited goal → provide only appropriate material → ask for a draft, analysis or artifact → inspect factual claims and the produced file or code → revise before use or sharing. The user remains responsible for the decision and any external action.
Chat on web, desktop and mobile, web search, file creation, memory, connectors, up to 5 projects; Sonnet 5 and Haiku 4.5
Pro
$20/month, or $17/month billed annually ($200)
At least 5x Free usage, Opus 5.5, Research, Claude Code, Claude Design, Slides and Docs
Max 5x / Max 20x
$100 / $200 per month
5x or 20x Pro usage, priority at busy times; monthly billing only
Team
$25 per seat monthly or $20 annually; Premium seat $125 or $100
2–150 people, SSO, central billing, no training on your content by default
Enterprise
$20 per seat per month (annual) plus usage at API rates
SCIM, audit logs, custom retention
Usage resets in five-hour windows, and paid plans add weekly caps. Chat, desktop, mobile and Claude Code all draw from one pool. When you hit the limit you can wait, upgrade, or switch on usage credits billed at API rates; turn those on deliberately. The top model, Fable 5.1, isn't on Free and uses extra credits on Pro. There's nothing to install or maintain; the admin work is choosing the plan and the privacy setting.
When to pay: stay on Free until the limits stop you several times a week, then take Pro. Choose Team as soon as two or more people use it for work: it doesn't train on your content by default, and billing is central.
Restrictions
The help centre identifies different interfaces and account features, while plan features can change. This review did not sign in, upload files, connect an app, use memory or establish the data terms for a particular account or organization.
First project
Documentation-based suggestion — not run here.
Read first project
Turn a made-up meeting transcript into minutes and an action list.
Input: a one-page fictional transcript of a team meeting with about five decisions and six tasks. Write it yourself, or ask Claude to invent one.
Sign in at claude.ai. Free is enough.
Open Settings → Privacy and set Model Improvement the way your organisation wants.
Paste the transcript and ask: "Write minutes with the decisions, then a table of actions with owner and due date. Write 'not stated' where the transcript doesn't say."
Ask for the action table as an Excel file and open it.
Check every action against the transcript. Tell Claude about one wrong or missing item and let it fix the file.
Done when: every action, owner and date in the spreadsheet traces back to a line in the transcript, and nothing is invented.
Links & evidence
Documentation-based · not hands-on · Sep 23, 2026
Rests on Anthropic's own pages read on 2026-09-23: claude.com/pricing (plans, features, models), the Max plan help article, the 16 September 2026 announcement merging Cowork into Claude, the model-training privacy article and the desktop install article. Nothing was run or tested for this review. Not verified: whether the Model Improvement switch is preselected at sign-up, exact usage allowances per plan (Anthropic publishes none), and EUR prices for German buyers.
Claude plans and pricing · Sep 23, 2026 — Free, Pro ($20 or $17/month annual), Max from $100, Team $25/$20 and $125/$100 per seat, Enterprise $20/seat plus usage; features per plan; usage pool shared with Claude Code.
What is the Max plan? · Sep 23, 2026 — Max 5x $100 per month and Max 20x $200 per month.
Claude Cowork and chat are now one Claude · Sep 23, 2026 — Merge announced 16 September 2026; rolling out to Pro and Max first, Team and Free later, Enterprise with 30 days notice.
Is my data used for model training? · Sep 23, 2026 — Consumer plans: chats and coding sessions train Claude only if Model Improvement is allowed; incognito chats never; setting location.
Install Claude Desktop · Sep 23, 2026 — Desktop app on macOS, Windows and Linux (beta); Claude Code and Cowork on paid plans only.
Further review notes & alternatives
Alternatives and trade-offs
ChatGPT wins when your team already uses it or you need its €8 Go tier; switch off "Improve the model for everyone" first.
Gemini wins when you live in Google Docs, Gmail and Drive.
Open WebUI wins when data must stay on your own hardware and someone can run a server.
Draft documents, inspect uploaded files, search with sources, and run small code artifacts.
Choose it for low-friction browser-based analysis and writing.
Its live Artifacts provide a concrete place to execute small code snippets.
Anthropic · free-tier · cloud
browser chat
document drafting
pdf analysis
csv parsing
image interpretation
web search integration
live code execution
talk-think
draft-docs
write-code
data-analytics
View sourced imageOfficial Claude web-chat interface screenshot from its getting-started guide.
Claude official product documentation
Owner-requested editorial use of public official interface screenshots, 15 September 2026. Copyright retained by the original publisher; attribution preserved. No open licence or hands-on course execution is asserted. Source asset: https://downloads.intercomcdn.com/i/o/lupk8zyo/1916208578/2cf2ea52f1f884084b57983a8805/image.png. PNG conversion preserves the complete frame.
