Claude review: The AI assistant that actually reads the whole document

Claude is a family of large language models built by Anthropic, wrapped in a chat app you can use on the web, on desktop (Mac and Windows) and on iOS and Android. Under the hood it is the same idea as every other assistant: you type, it answers. What makes Claude different is everything that grew around the chat box over the last two years.

Here is what you actually get today:

• Chat with file uploads: PDFs, Word files, spreadsheets, images and code.
• Artifacts: a side panel where drafts, code and small apps live while you iterate.
• Claude Docs, Slides and Design: built-in creation tools available even on the Free plan.
• Claude Code: an agentic coding tool for your terminal and editor, included with paid plans.
• Claude in Chrome and Microsoft 365: Claude inside your browser, Word, Excel and Outlook on paid plans.
• Memory, connectors and Skills: persistent context, links to your apps, and reusable task recipes.

The pitch has quietly shifted from "a nicer chatbot" to "a coworker that can finish tasks". You can hand off and schedule tasks on Pro, which is something we did not expect to use and now use weekly for recurring research summaries.

Anthropic names its models by size, and the naming finally makes sense once you have used them for a week. Every current model handles text and images, writes in many languages and can use tools.

Our honest take: Sonnet is the right default for 80% of what most people do. We switch to Opus for gnarly debugging, contract reviews and anything where a wrong answer is expensive. Fable is impressive, but it eats your allowance quickly, so we save it for the genuinely hard stuff.

The current models support a context window of up to 1 million tokens, which Anthropic puts at roughly 555,000 words. In practice, what fits in a single chat depends on the model and your plan, so treat the headline number as a ceiling rather than a promise.

This is the reason most people end up paying for Claude, and it is the reason we did. Claude writes like a careful editor rather than a marketing intern. It picks up on tone, keeps sentences varied and, crucially, follows your style instructions for the whole document instead of drifting back to generic AI voice after three paragraphs.

What we use it for

• First drafts from a messy outline and a pile of notes.
• Line edits that tighten prose without flattening your voice.
• Restructuring: "turn this 4,000-word brain dump into a clear brief with headings."
• Tone shifts: the same email rewritten for a CFO, a customer and a friend.

Where it still needs a human

Claude is good, not magic. It will occasionally hedge more than you want, and when it does not know something it can still produce a confident-sounding sentence. Our rule is simple: Claude drafts, a human checks facts, numbers and quotes. If you are shopping around specifically for writing, our roundup of the best AI writing tools compares it with Jasper, Grammarly and others.

A small tip that made a big difference for us: put your style guide in a Project once (banned phrases, preferred spelling, sentence length) instead of pasting it into every chat. The quality jump is noticeable.

If writing is why people pay, long-context work is why they stay. We threw a 212-page annual report, a stack of interview transcripts and a sprawling legal agreement at Claude, and the results were consistently better than we expected.

The trick is not just that the files fit. It is that Claude actually uses the middle of the document. Ask about a footnote on page 140 and it finds it, quotes it, and tells you how it contradicts the summary on page 6. Older assistants (and plenty of current ones) tend to skim the start and the end and improvise the rest.

How we get the best results

• Ask Claude to quote the exact passage it relied on. It makes checking fast.
• Split giant jobs into stages: summarize, then extract, then analyze.
• Use a Project for documents you keep coming back to, so you are not re-uploading them.
• Remember that huge uploads drain usage much faster than short chats.

That last point matters more than any spec sheet. A 1M-token context is wonderful, but every message in a long chat re-reads the whole conversation, which is exactly how you burn through a usage window before lunch.

These features sound like checkbox items on a pricing page. In daily use they are what separate Claude from a plain chatbot.

Artifacts are the one we would miss most. When Claude writes a document, a chunk of code or a small interactive tool, it opens in a side panel instead of being buried in the chat. You can keep asking for changes and watch the artifact update, then copy, download or share it. We have built quick calculators, a content calendar and a pricing comparison widget this way without opening an editor.

Projects are folders with a brain. Drop in reference files and a set of instructions, and every new chat inside that Project starts with that context. Free users get up to five, which is enough to see whether they click for you.

Memory lets Claude carry context between conversations, and you can view and edit what it remembers. Connectors link Claude to tools like Google Drive, Gmail, calendars and many third-party apps, so you can ask "what did the client say about the deadline last week?" without copying anything in. They work well, but think twice before connecting sensitive inboxes on a personal account.

We did not plan to write a big coding section in a general review, but leaving it out would be dishonest. Claude Code is included on every paid plan, and for many developers it is now the main reason to subscribe.

It runs in your terminal or editor, reads your repository, plans changes across multiple files, runs your tests and asks before doing anything destructive. On our own codebase it handled refactors, migrations and annoying "why does this only fail in production" bugs with a level of care that genuinely surprised us. It also reads instructions from a project file, so you can tell it your conventions once.

• Great for: multi-file refactors, test writing, debugging with real logs, explaining unfamiliar code.
• Watch out for: long agent runs eat the same usage pool as your chats.
• Compared with: editor-first tools like Cursor and GitHub Copilot, Claude Code feels more like delegating a task than autocompleting a line.

If you code every day, this alone can justify Pro, and it is the strongest argument for Max.

Here is the thing that trips up almost every new subscriber. Claude does not give you a fixed number of messages. Instead, usage resets on a rolling five-hour window, and paid plans also have weekly limits on top. How far you get depends on how long your chats are, which model you pick and which features you use.

According to Anthropic, Pro gives you at least five times the usage of Free per session, and Max gives you 5x or 20x the usage of Pro. Chats, the desktop app, mobile and Claude Code all draw from the same pool, which is great for simplicity and terrible if you forget a long coding session is running.

How to stretch your allowance

• Start new chats for new topics instead of one endless thread.
• Use Sonnet by default and switch to Opus only when you need it.
• Keep reference material in Projects rather than re-pasting it.
• Check Settings, then Usage to see where you stand before a big task.

When you do hit the wall on a paid plan, you can wait for the reset, upgrade, or turn on usage credits billed at API rates. It is fair, but it is not "unlimited", and nobody should buy Pro thinking it is.

People search for "Claude daily limit" or "Claude monthly limit", so let us be precise about what exists. There is no daily cap. Anthropic uses two stacked limits instead, and it does not publish a fixed number of messages for any plan. The official usage and length limits page describes the same structure.

• Session limit (every five hours): applies to every plan, Free included. Your allowance refills five hours after you start using it, not at midnight.
• Weekly limit: applies to paid individual plans and covers all models. It resets at a fixed day and time assigned to your account, and you can see your next reset under Settings, then Usage.
• Monthly limit: none is published. Anthropic does say it may apply weekly and monthly caps, or limits on certain models and features, at its discretion to keep access fair.

Limits at a Glance

One caveat worth knowing: Anthropic states the 5x and 20x multiples for the per-session allowance. It does not say the weekly limit scales by the same factor, so do not assume it does. Third-party message counts you will see online are estimates, and Anthropic itself says the real number depends on how you use Claude.

Team and Enterprise

Seat-based Team and Enterprise plans give each seat its own allowance, and admins can buy extra usage credits. Usage-based Enterprise plans work differently, because you are billed for what you consume rather than hitting a fixed cap. Anthropic does not list per-seat numbers publicly, so ask sales if limits matter for your team.

What Eats Your Allowance Fastest

Everything draws from one shared pool: claude.ai, the desktop app, mobile and Claude Code. Four things drain it fastest:

• Long conversations, because the whole thread is processed again with each message.
• Opus, which uses meaningfully more than Sonnet.
• Big attachments and tool-heavy tasks such as research, web search and connectors.
• Higher effort levels and extended thinking.

Content stored in a Project is cached and counts less when you reuse it, which is one more reason to stop re-pasting the same PDF.

What to Do When You Hit a Limit

You have three options: wait for the reset, upgrade (the new plan's limits apply right away), or switch on usage credits so work carries on at pay-as-you-go rates instead of stopping. Our rule of thumb is simple. Stay on Free while you are testing, move to Pro when the five-hour window runs out before your workday does, and only pay for Max if you hit the weekly limit regularly, which in practice means daily Claude Code use.

Worth reading before you paste in anything confidential. On the personal plans (Free, Pro and Max), Anthropic lists model training as opt-out, which means you should check your privacy settings and switch it off if you do not want your chats used. Team and Enterprise plans do not train on your data by default.

Claude also offers incognito chats that are not saved to your history or memory, which we use for anything sensitive on a personal account. For business use, the Team and Enterprise plans add single sign-on, admin controls and, on Enterprise, audit logs, custom data retention and a HIPAA-ready offering.

If your company is writing rules for AI use, the NIST AI Risk Management Framework is a practical, vendor-neutral starting point.

After months of daily use, the pattern is clear.

Claude is a great fit if you:

• Write for a living, or write a lot for work.
• Regularly deal with long PDFs, contracts, research or transcripts.
• Write code and want a capable agent rather than just autocomplete.
• Care about tone and want output you do not have to de-robot.

Look elsewhere if you:

• Mainly want to generate images. Claude does not create pictures natively.
• Want a single app with image, video and voice features all built in.
• Need cited answers for every claim by default, where a research engine fits better.

The most common question we get is whether to pick Claude or ChatGPT. Our ChatGPT review covers the other side, and the head-to-head ChatGPT vs Claude comparison settles it task by task. If you want to know how we scored this, see how we review AI tools.

Prices above exclude tax and were taken from Anthropic's pricing page in October 2026. Anthropic changes plans fairly often, so confirm before you subscribe. Our advice: start on Free, move to Pro the first week you hit a limit twice, and only consider Max if you use Claude Code daily or live in Claude all day. Pro can be cancelled any time, and your chats and Projects stay on your account if you drop back to Free.

Claude is the assistant we trust with work that matters. It writes better than anything else we have tested, it actually reads the long documents you give it, and Claude Code turned it from a writing tool into a genuine engineering partner.

The catch is real: usage limits are fiddly, the best models sit behind a paywall, and there is no image generation. But for writers, analysts, researchers and developers, Claude Pro is the easiest $20 a month in our software budget. Start free, push it with a real document, and you will know within a day whether it is for you.