To use Claude, open it at claude.ai or in the desktop or mobile app, pick a model, and give it a real task from your week with the briefing you would give a new colleague: the goal, the background, the source material and the format you want back. Claude is Anthropic's AI assistant, and at work it performs best when you attach the actual files, say who the output is for, and ask for revisions instead of starting over.
This guide on how to use Claude covers the apps, plans and models, the features that matter at work, ten tasks with prompts to copy, and free ways to learn more.
Where to use Claude: web, desktop and mobile
You can use Claude in any browser at claude.ai, in a desktop app, or on your phone. Anthropic's download page lists desktop apps for macOS, Windows and ChromeOS, a Linux beta for Ubuntu and Debian, and mobile apps for iOS and Android. According to the Claude Help Center, you must be at least 18 and in a supported location to sign up.
Install the desktop app if you work with local files: it lets you give Claude access to a folder so it can read, organize and create files there. Anthropic is also merging chat and Claude Cowork into one conversation, starting with Pro and Max plans, per this Help Center article; if you still see separate "Chat" and "Cowork" options, you have the previous layout.
Claude plans: what is free and what you pay for
The prices below come from Anthropic's official pricing page as of September 2026, and they exclude tax.
| Plan | Price | Best for | What it adds |
|---|---|---|---|
| Free | $0 | Trying Claude on real tasks | Chat on web, desktop and mobile, web search, file creation, memory, connectors, artifacts, up to 5 projects |
| Pro | $20/month, or $17/month billed annually | Individual professionals | More usage, Opus models, Research, Claude Code, Claude in Chrome and Microsoft 365, Claude Design, Slides and Docs |
| Max | From $100/month | People who work in Claude all day | 5x or 20x the usage of Pro, higher output limits, priority access at busy times |
| Team | $25 per seat/month, or $20 billed annually (standard seat) | Teams of 2 to 150 people | Shared projects, central billing, single sign-on, admin controls, no model training on your content by default |
| Enterprise | $20 per seat/month billed annually, plus usage at API rates | Large organizations | Role-based access, audit logs, custom data retention, HIPAA-ready option |
A sensible path: start on Free with one real task, and upgrade when you hit usage limits or need Opus, Research or the Chrome and Microsoft 365 extensions.
Which Claude model should you choose?
Anthropic's models overview lists four current models:
- Claude Fable 5.1: for demanding reasoning and long-horizon agentic work; the slowest of the four.
- Claude Opus 5.5: for long-running agentic coding and knowledge work; the docs suggest starting here when unsure.
- Claude Sonnet 5.5: the best combination of speed and intelligence.
- Claude Haiku 4.5: the fastest model, good for quick, simple requests.
A fifth model, Claude Mythos 5.1, is limited to vetted cyberdefenders and life scientists, according to its official page, so it won't appear in your model picker.
Click the model name below the message box (at the top of the screen on mobile) to switch. Your plan decides the menu: Free includes Sonnet and Haiku, while Opus needs a paid plan (pricing). Fable 5.1 runs on pay-as-you-go usage credits on Pro, and Max plans can spend up to 50% of their weekly limits on it, per the Help Center.
A practical default: Sonnet for everyday drafting and analysis, Opus for long, multi-step work across several documents. Our guide to Claude Sonnet 5.5 for everyday work shows how to set it up for daily tasks.
The Claude features that matter at work
File uploads
Upload the source instead of pasting fragments. Claude reads PDF, Word, CSV, Excel, JSON, plain text and image files, and one chat accepts up to 20 files of up to 500 MB each, according to the upload guide. The same guide notes that Claude reads charts inside PDFs of 100 pages or fewer, while longer PDFs, up to 1,000 pages, are processed as text only.
Projects
A project is a workspace with its own chats, a knowledge base of files, and custom instructions, such as "answer as a financial analyst writing for executives." Free accounts can create up to five projects, paid plans expand project knowledge up to 10x through retrieval, and Team and Enterprise members can share projects, as described in What are projects? For recurring work, write the instructions once and every chat in the project starts with that context.
Artifacts, documents and decks
Ask for a document, chart, prototype or visual and Claude can create it as an artifact you keep editing in the conversation; every artifact gets a link that stays private until you share it, according to Anthropic's artifacts page. Claude can also produce spreadsheets with working formulas and presentations you can open in PowerPoint (Help Center).
Connectors
Connectors let Claude read from and act in apps like Google Drive, Gmail, Slack or Linear, and Claude only reaches what your own account can access, according to the connectors guide. On Team and Enterprise plans an Owner must enable each connector first. Start with read tasks, like search and summarize, before you let Claude send or change anything.
Memory and Research
Claude can carry context from past conversations into new ones, and paid plans can ask it to search previous chats (memory guide). Research, on paid plans, runs searches that build on each other across the web and your connected apps and returns answers with citations (Research guide).
10 work tasks to try in Claude, with example prompts
Copy these and replace the brackets with your details.
| Work task | Example prompt | Why it works |
|---|---|---|
| Summarize a report | "I'm attaching our Q3 customer survey. Summarize it for our VP of Sales in 5 bullets, then list 3 risks it mentions but doesn't emphasize." | Names the reader |
| Reply to a tough email | "Draft a reply to this client email [paste]. Calm and firm: we can't move the deadline, but we can deliver the first module early. Under 150 words." | States the limit and the offer |
| Meeting notes to actions | "Turn these meeting notes [paste] into a table with owner, task and due date. Flag any task with no owner." | Ready for your tracker |
| Analyze a spreadsheet | "This CSV has 12 months of support tickets. Find the three fastest-growing categories, chart the trend, and explain the likely cause." | Uses file analysis |
| Build a presentation | "Turn this project update [attach] into a 10-slide deck for non-technical leaders. One idea per slide, with speaker notes." | Sets audience and length |
| Research a market | "Search the web for project management tools for small agencies. Compare pricing model, integrations and learning curve in a table, with sources." | Asks for sources |
| Rewrite for customers | "Rewrite this release note [paste] for customers. Keep every feature, drop ticket numbers, and add why each change matters to them." | Says what to keep and cut |
| Document a process | "Interview me one question at a time about how we onboard clients, then write a step-by-step SOP with a checklist." | Claude gathers the context |
| Practice a negotiation | "Act as a procurement manager at a mid-size retailer. I'm selling [product]. Push back on price like a real buyer, then critique my answers." | Role play |
| Write a formula | "In Google Sheets, column A has dates and column C has deal values. Write a formula that sums C for the current month and explain it." | Exact inputs |
Prompts from other assistants transfer too: our ChatGPT prompts for work adapt to Claude with small edits.
How to write better Claude prompts: Anthropic's own advice
Anthropic's prompting best practices target developers, but these techniques work just as well in the chat window:
- Be clear and direct. Treat Claude like a very capable new hire who doesn't know your norms yet. If a colleague with no background would be confused by your prompt, Claude will be too.
- Explain why. "Keep it short because my manager reads it on her phone" works better than "keep it short," because Claude can generalize from the reason.
- Show examples. Examples reliably steer format and tone, and the docs recommend 3 to 5 of them for best results (best practices).
- Separate the parts. Wrap instructions, context and input in tags such as
<instructions>,<context>and<input>so Claude doesn't mix them up. - Give Claude a role. One sentence, like "You are a CFO reviewing a budget request," focuses tone and criteria.
- Put long documents first. Place long inputs at the top and your question at the end; Anthropic reports this can improve response quality by up to 30% in its tests (best practices).
- Say what to do, not what to avoid. "Write in flowing paragraphs" beats "don't use bullet points."
- Draft, review, refine. Ask for a draft, have Claude review it against your criteria, then revise. The docs call this self-correction.
Here is how those techniques combine in one work prompt:
<context>
I lead customer success at a B2B software company. Renewals dropped last quarter.
</context>
<input>[attach the churn report]</input>
<instructions>
You are a customer success analyst. Give the three likeliest reasons for the drop,
with evidence for each, and one action per reason. For my VP: short, no jargon.
</instructions>Privacy, data and limits to check before you paste work into Claude
- Personal plans. On Free, Pro and Max, you choose in Privacy Settings whether your chats can be used to improve Claude, and raw content from connectors is not part of that data, according to the Anthropic Privacy Center.
- Business plans. Team and Enterprise do not use your content for model training by default (pricing).
- Incognito chats. The ghost icon starts a chat that isn't saved to your history or memory and isn't used for training, though it is retained for 30 days by default (incognito guide).
- Company policy. If your employer provides a Team or Enterprise account, use it for work documents.
- Usage limits. The Free plan has a session-based limit that resets every five hours (Help Center).
- Accuracy. Claude can write answers that look right but are wrong, and Anthropic advises against treating it as a single source of truth (Help Center). Check numbers, names, quotes and anything legal or financial.
Claude AI courses and training: where to learn more
You don't need a paid Claude AI course, but structured practice helps. Anthropic's Claude Academy offers free courses, and its course catalog lists the lessons and durations below; Anthropic also maintains an interactive prompting tutorial on GitHub with 9 chapters of exercises.
| Resource | Format | Best for |
|---|---|---|
| Claude 101 (Claude Academy) | Free, 13 lessons, about 2.5 hours | Learning the product |
| AI Fluency: Framework and foundations (Claude Academy) | Free, 14 lessons, about 4 hours | Anthropic's 4D framework for working with AI |
| AI capabilities and limitations (Claude Academy) | Free, 13 lessons, about 3.5 hours | What models can and can't do |
| Anthropic's prompting tutorial on GitHub | 9 chapters with exercises | Technical users |
| 4Geeks Prompt Engineering Course for Beginners | Free interactive exercise, about 5 hours | Practicing prompts with instant feedback |
Our free prompt engineering course teaches the techniques from scratch: you write prompts and see the output in real time, each attempt is validated instantly, an AI assistant helps when you get stuck, and your progress is saved. For the wider picture beyond Claude, read our guide on how to use AI at work.
To see which other tools fit your work, browse our map of AI tools.
Your next step: run one recurring task through Claude this week
- Pick a task you repeat weekly, such as a status report or client follow-up.
- Create a project for it with instructions on audience, tone and format, plus two or three strong past examples.
- Run this week's version and ask Claude to review its draft against your instructions.
- Add whatever you had to fix to the instructions, so next week starts closer to done.
If you'd rather build this skill with a mentor, our AI Fluency program (separate from Anthropic's free course with a similar name) runs 4 weeks, requires no coding, and includes 1:1 mentorship, Rigobot, our 24/7 AI tutor, and a certificate. For a deeper track, our 6-week Applied AI Course for professionals comes with a certificate recognized by the Florida Department of Education. See both side by side on our program comparison page.