Meet Claude. Try a useful first task.

Getting started

Section one is getting started. Your account and your device shape which tools you see, so we begin there: the web, desktop, and mobile versions; what changes between a personal plan and a workplace account; a first look at the conversation screen; the difference between a quick answer and a task; and three basic moves you will use every day: setting the permission mode, attaching a file, and reading a response.

Choose a device

Claude runs in three places. On the web, at claude.ai in your browser: good for chat, uploading an approved file, or following a task. On the desktop, through the official Claude Desktop app: this is where Claude can work with approved local files and desktop tools. And on mobile, through the official iOS or Android app, shown on the right: good for questions and for continuing work away from your desk. Use the same intended account and workspace on each device. Install only the official app or an organization-managed copy. Features vary by operating system, plan, administrator settings, and rollout.

Know your account

A paid personal plan and a workplace account have different controls, and it helps to know which is which. You may choose your prompt and attachments, your personal instructions and appearance, the models and effort levels available to you, and connector use within your permissions. Your organization may decide which models and tools appear, which services you can connect, which sites and actions are allowed, and the retention, sharing, and usage policies. So before you add any business information, check the signed-in email and workspace. And remember: a missing setting can reflect policy, not a problem with the app.

Find five screen functions

Let's find five functions on the screen. Marker one, New chat, at the top of the sidebar: start a fresh topic and give the task a clear name. Marker two, the message box: type your request and include the facts Claude needs. Marker three, the plus button: add an approved file or choose tools for this task. Marker four, the model menu, next to the send button: this is where the model and effort controls live. Marker five, history and settings: Recents in the sidebar takes you back to prior work, and your name at the bottom opens your account controls. Take a moment to locate each of these on your own screen.

Chat and Cowork

Depending on your account, you may see one of two layouts. On the left, the older layout has separate options: Chat, useful for discussion and quick answers, and Cowork, which takes on work involving tools, files, and multiple steps. On the right, the newer experience has one message box that can answer questions and take on tasks in the same conversation. The unified experience is rolling out gradually, so your screen may still have the selector. Either way, the useful distinction is the same: are you asking for an answer, or authorizing work that can affect files or services? Describe the outcome and the boundaries.

Quick answer or task

The wording of your request should make your expectation clear. A quick answer looks like this: "Explain the difference between a prompt and a model in three sentences." No files or connected services are needed. A task with a boundary looks like marker one: "Use these fictional notes to create a one-page staff briefing. Save a new copy. Ask before changing any source file." Then, at marker two, check the plan Claude shows and confirm it matches what you asked for. At marker three, approve only what you intended, and review the result before you use it. A specific outcome reduces ambiguity, and a stated boundary protects your files. A task can keep running in the cloud, but work with files or apps on your computer needs the desktop app open and connected.

Permission modes

That approval card comes from a setting in the message box that controls how much Claude checks with you. In Manual, which is the default, Claude asks before it takes an action, and you choose whether to allow each one. In Auto, Claude keeps working without stopping to ask about each step, and automated safety checks run before it acts. The setting applies to the whole conversation, and you can change it, stop Claude, or redirect it at any time. For anything with real consequences, such as sending messages or changing important files, stay in Manual and review what Claude does. On Team and Enterprise plans, your administrator controls whether automatic approval is even available.

This slide has the camera, so it's worth a screenshot.

Attach an approved file

Attaching a file is the most common first action. Click the plus button in the lower left of the message box and choose "Add files or photos," at marker one. You can also drag a file into the chat window or paste an image. The attached file appears as a chip above your message, at marker two. Then write your request, at marker three: say what you want done and what must be preserved. Claude accepts common documents, such as PDF, Word, text, CSV, and spreadsheets, and common image formats, up to twenty files per chat. Two habits: attach a copy, not your only version, and never attach something you are not cleared to share.

Read and review a response

Every reply comes with controls worth knowing. When Claude searched the web, you will see a line saying so, at marker one, and citations inside the answer, at marker two; open those before relying on a fact. Below the reply are the actions at marker three: copy the text, give a thumbs-up or thumbs-down, or retry for a fresh answer. And marker four, Edit on your own message: change the wording and resend, rather than arguing with a bad answer. One caution: a thumbs-down report can send the conversation to Anthropic for review, so do not rate a conversation that contains sensitive material without checking your organization's policy first.