- Guides
- Tool guides
- How to use Claude
How to use Claude
What Claude is, how to set up projects and connectors, and the habits that get better answers on the first try.
We have no affiliate deal with Anthropic, and nobody paid to appear on this page.
Claude is Anthropic's AI assistant. You can use it free at claude.ai, in the desktop app for Mac and Windows, or on iOS and Android 1. For work, the three features that matter most are Projects (saved context for recurring tasks), connectors (Gmail, Drive and other tools) and long-file handling. This guide covers each, with prompts you can copy.
What Claude is
Claude is both the app and the family of models behind it. The current models are Opus 5.5 (Anthropic's default for most work), Sonnet 5.5 (faster), Haiku 5.5 (fastest) and Fable 5.2 (the most demanding reasoning) 16. You pick one from a menu in the chat, and the free plan uses Sonnet and Haiku 2.
Claude is strongest at reading and writing: long reports, contracts, transcripts, spreadsheets and code. On paid plans the newest models hold 1 million tokens in one chat 15, enough for a stack of documents at once. It does not generate photos or illustrations; it builds charts, diagrams and interactive visuals instead 13.
You must be 18 or older to use it 1. The free plan includes web search, file creation, memory, connectors and up to five projects 2, which is enough to follow every step below except Research, Claude in Chrome, tasks and computer use, which need a paid plan.
Your first conversation
The quality of the answer tracks the quality of the brief. Anthropic's own first tip is to be specific 1.
- Sign in at claude.ai.Or download Claude Desktop or the mobile app and sign in there with the same account.
- Pick the model and effort.Click the model name next to the send button. Low or Medium effort suits routine tasks; High offers the best balance of quality and speed; Extra high and Max are for hard problems 3.
- Give the whole brief in one message.Say who you are, what you need, what to use, and what the output should look like. Attach the files in the same message.
- Use the + button.Click + in the lower left, or type
/, to add files, connectors, Research and other tools 1. - Push back.Ask follow-ups, point at what is wrong, and ask for a second version. The first answer is a draft.
A first prompt that works for most business tasks:
I run a 12-person design agency. Attached are last quarter's three client feedback reports. List the five complaints that come up most, quote the exact line behind each one, and suggest one fix per complaint ranked by effort. Keep it under 300 words and use a table.
Habits that get better answers
- Batch related questions into one message instead of five short ones. You get a more joined-up answer and use less of your limit 12.
- Ask it to ask first. End a big request with "Ask me anything you need before you start."
- Ask for the source of every number. Have Claude name the file, page or row behind each figure, so a wrong one is easy to spot.
- Talk instead of typing when the context is long. A two-minute spoken brief often carries more detail than a typed one.
Projects for recurring work
If you do a task more than twice, give it a project. A project keeps instructions and reference files together, and every chat inside it starts with that context 4.
How we use them: our editors used to re-explain the same background to Claude at the start of most writing sessions. Moving that work into one project with the source documents and templates loaded cut the setup time and made first drafts more consistent.
- Create it.Click Projects in the sidebar, or go to claude.ai/projects, then + New Project. The name and description are for you; Claude does not see them 4.
- Set the instructions.Click Set project instructions and write how Claude should work in this project: tone, format, rules. They apply to every chat in it.
- Add knowledge.Click + on the project page to upload documents, text or code. Each project file can be up to 30 MB 6. On paid plans, Claude switches to a retrieval mode when the knowledge nears the context limit, so the project can hold more 4.
- Work in project chats.Each project has its own memory. To keep it separate from the rest of your chats, turn off Use account memory on the project page 4.
Instructions we would put in any project that handles numbers:
Every number you give must name the file and the row or page it came from. If figures don't reconcile, list them under "Exceptions" instead of forcing them to fit. If data is missing, say so. Never estimate a figure without labelling it as an estimate.
Free accounts can create up to five projects. On Team and Enterprise you can share a project with colleagues as "Can view" or "Can edit" 4.
Projects or skills? A project is background knowledge that is always loaded in its chats. A skill is a saved procedure Claude pulls in only when a task needs it, in any chat 11. Use a project for a client or topic; use a skill for a repeatable how-to, such as your report format. We compare Projects with Gems and GPTs in Gemini Gems vs Custom GPTs vs Claude Projects.
Connectors: Gmail, Drive and more
Connectors let Claude read from and act in the tools you already use. The Gmail, Google Calendar and Google Drive connectors are available to all users on the web and desktop apps 5; on Team and Enterprise an owner has to switch them on first.
- Open the menu.In a chat, click +, hover over Connectors and toggle on the ones you want 5.
- Approve access.Sign in to the Google account you want Claude to use and grant access.
- Name the tool in your prompt."Search my Gmail..." or "Find the Q3 deck in Drive..." tells Claude where to look.
What each one does, per Anthropic: in Gmail, Claude can search and read mail, draft, manage labels, and send, reply or forward. It reads attachment details but not their content. In Calendar, it can view and manage events and find free time. In Drive, it can search, read, upload, share and move files 5. By default, Claude asks for your approval before every send, share, move or trash.
Claude can also edit Google Docs, Sheets and Slides live in a pane beside the chat, on the web in Chrome or in Claude Desktop 5. Review those edits as you would your own, because they land in your file. Paid plans add Claude for Microsoft 365 2.
Search my Gmail for messages from Acme in the last 30 days. List every open request with its date and who asked. Then draft one reply that answers each request. Do not send it.
Each connector you leave switched on adds to your usage, so turn off the ones a task doesn't need. More in Claude usage limits, explained.
Files, long documents and research
Claude handles big uploads well. In a chat you can attach up to 20 files, each up to 500 MB, and PDFs up to 1,000 pages 6. For PDFs up to 100 pages, Claude reads charts and images as well as text; longer PDFs are read as text only.
Supported documents include PDF, DOCX, CSV, TXT, HTML, RTF, EPUB, JSON and XLSX. Spreadsheets need code execution and file creation turned on 6.
Attached are our MSA and the client's redlines. List every change that shifts risk or cost to us, quote the clause and page, and rate each one high, medium or low. End with the three points to push back on first.
Research is for questions that need many sources. It is on paid plans only. Turn on web search, click +, and select Research; a blue indicator shows it is on 7. It runs on Opus 5.5 whatever model you picked, and it can search your connected Google Workspace too. Start it on the web or desktop: on mobile, Research searches only the web.
Tip: ask Research for a table of claims with the source next to each one. It makes the fact-check pass much faster.
Claude in Chrome and on your computer
Claude can also work in your browser and on your desktop, not just in the chat window. These features are newer, and Anthropic is clear that they carry risk.
Claude in Chrome
Claude in Chrome lets Claude read, click and navigate websites alongside you. It is on all paid plans and works only in Google Chrome on a computer, not in other Chromium browsers or on mobile 8.
- Install it.Find Claude in Chrome in the Chrome Web Store and click Add to Chrome.
- Sign in and pin it.Sign in with your Claude account and pin the extension from the puzzle-piece menu.
- Grant permissions.The debugger permission is what lets Claude control the browser 8.
Our setup advice: run it in a separate Chrome profile with no saved logins, so a confused agent has nothing valuable within reach. Start with one boring, low-stakes task on one site, tell Claude exactly when to stop, and watch the whole run before you trust it with more.
Tasks, scheduling and computer use
Claude can now take on multi-step tasks from the same message box as chat. Anthropic folded its Cowork agent into the main app ("Chat and Cowork share one home"), and Claude decides whether a request needs a quick answer or a task. Cowork is on paid plans only; the combined app is rolling out to Pro, Max and Ultra, and task sessions run in the cloud 9. Type /schedule in a task, or click Scheduled in the sidebar, to run it on a schedule even when your computer is off.
Computer use lets Claude operate apps on your Mac or Windows machine when no connector exists. It is in beta on Pro, Max and Ultra. Turn it on in Claude Desktop under Settings, Capabilities (or Settings, System, under This computer), Enable computer use 10. Claude asks before opening each app and blocks trading and crypto apps by default. Anthropic advises against using it with banking, healthcare or government apps.
Which plan you need
Free covers the basics. Pro at $20 a month is the plan most people who use Claude for work end up on 2.
| You want to | Plan | Price |
|---|---|---|
| Chat, upload files, search the web, use connectors and up to 5 projects | Free | $0 2 |
| Use Opus, Research, Claude Code, Claude in Chrome and Microsoft 365 | Pro | $20, or $17 billed yearly |
| Let Claude operate apps on your computer | Pro, Max or Ultra, in beta | From $20 10 |
| Use Claude all day, with Fable included | Max or Ultra | $100 or $200 |
| Share projects with colleagues | Team | $25 a seat, or $20 billed yearly |
On Free, Pro and Max, Anthropic trains on your chats only if the model training setting in Privacy Settings is on. With it on, data is kept for five years; with it off, the 30-day retention period applies. Team and Enterprise are not covered by that change 14.
For the full plan breakdown, see Claude Pro vs Max, and Sonnet vs Opus. If you are weighing Claude against another assistant, start with ChatGPT vs Claude or Claude vs Gemini.
Related
New features, explained the week they ship
195,000 readers get Uncover AI's practical AI briefing. We show you what is new and how to use it in your work.
Sources
- Get started with Claude
- Claude pricing
- Change the model, effort and thinking settings
- How can I create and manage projects?
- Use Google Workspace connectors
- Upload files to Claude
- Use research on Claude
- Get started with Claude in Chrome
- Get started with Claude Cowork
- Let Claude use your computer in Cowork
- What are skills?
- Usage limit best practices
- Can Claude produce images?
- Updates to Anthropic's consumer terms
- How large is the context window on paid Claude plans?
- Claude models overview