How to Actually Use Claude: A Field Guide to Every Mode, Feature, and Extension

If you only open Claude, type a question, and read the answer, you are probably using only part of what it can do. That is not a criticism. It is simply how these tools tend to evolve. The chat window gets all the attention, then more powerful features get added around it. Most people never find those features because no one gives them a clear tour.

A helpful cheat sheet has been circulating that lays out the main pieces in one place: the different ways to use Claude, the core features that make it useful for real work, and the extensions that connect it to the rest of your digital life. This guide builds on that idea. It explains what each piece does, where the simple descriptions need more context, and how to start using the parts you may have ignored.

The three modes: where you actually type

Chat: the default place most people start

Chat is the plain conversational interface most people mean when they say they asked Claude something. You type, Claude replies, and the conversation continues from there. It is still the right place for a lot of work, including quick questions, first drafts, brainstorming, and explanations. There is nothing wrong with spending most of your time here.

Two settings are worth turning on before you do anything else:

• Extended thinking, which lets Claude reason through a problem step by step before answering instead of firing off the first plausible response. It matters most on anything with real logical structure: math, multi-step planning, debugging, or a question where the obvious answer might be wrong.
• Web search, which lets Claude look things up instead of relying only on what it learned during training. Anything time-sensitive — current events, prices, “what’s the latest version of X” — benefits from this being on.

While you are in the settings area, check the Privacy section. That is where you can control whether your conversations may be used to help train future models. There is not one right answer for everyone, but it is a setting worth understanding.

Cowork: the desktop app for heavier tasks

Cowork is a separate desktop application designed for longer, multi-step work. Instead of having you feed Claude one instruction at a time, it lets you hand over something closer to a small project. For example, you might ask it to reconcile invoices, organize research notes, or draft a report from several documents. Claude can work through the task using available tools, then check in when a human decision is needed.

The most useful setup step is giving it a dedicated work folder. That gives Cowork a place to read from and write to, so it does not start from scratch every time. A short markdown file can also help. Use it to describe your role, your preferred writing style, your recurring projects, and how you want work delivered. That small bit of context can make the tool feel much less generic. Cowork can also be reached from the Claude mobile app, which makes it possible to check on tasks while you are away from your desk.

Code: an agentic coding tool, not just a terminal feature

Code is built for developers who want to delegate real coding work. It can read a codebase, write and edit files, run tests, and help debug problems. It is easy to think of it as only a command-line tool, but it is broader than that now. You can use it in the terminal, inside VS Code and JetBrains IDEs, through the desktop app, and from the mobile app.

If you are not a developer, you can safely skip Code for now. Learn Chat and the core features first. If you are a developer and have not tried it yet, point it at a real repository and give it a small task. The difference between an AI that suggests code and an AI that can run tests, see what breaks, and fix the problem is larger than it sounds.

The core features: what makes Claude more than a search box

Projects: a workspace with memory

A Project is a dedicated space for one recurring job, client, topic, or body of work. It is more than a single conversation. You can give it reference documents, such as style guides, past work, specifications, or research notes, and each conversation inside the project can draw from that material. Projects are available on the free plan, with paid tiers offering more capacity and larger context windows.

The best rule is to create one project for each recurring task or deliverable. Do not make one giant project for everything. A project for a Q3 marketing campaign, loaded with the brand brief and past campaign copy, will usually work better than a single project that tries to hold your entire job. Give it clear standing instructions at the beginning, including tone, preferred format, and what good work looks like for that situation.

Artifacts: things you build with Claude

When you ask Claude to build something, such as an interactive calculator, a chart, a small web app, or a document, it can create a standalone Artifact instead of only describing the result. You can view it, revise it, reuse it, and publish it when it is ready. That makes Artifacts useful when the goal is not just an answer, but an actual working output.

In practical terms, it is the difference between having Claude explain how a budgeting spreadsheet could work and having Claude produce the spreadsheet itself.

Research: asking Claude to dig deeper

Research mode is useful when one quick search is not enough. It fits questions that require comparison, source checking, trend analysis, or a broader scan of current information. Claude can review many sources, compare what it finds, and return a cited report. This can take several minutes, but that extra time is the point. The goal is deeper work, not a fast surface-level answer.

The quality of the result depends heavily on the prompt. A vague request usually produces a vague report. A strong request explains what you already know, what decision you are trying to make, what sources matter, and what kind of final answer would be useful.

The extensions: plugging Claude into everything else

This is the layer many people skip, even though it is where some of the most useful workflow improvements live. Skills, Connectors, and Plugins all serve different purposes. Together, they help Claude move beyond a chat window and work more directly with the tools, instructions, and information you already use.

Skills: saving the way you want work done

A Skill is a reusable set of instructions for a specific task. Instead of explaining your preferences every time, you can save the process once and use it again. Built-in skills can help with file creation, including Word, Excel, PowerPoint, and PDF outputs. Custom skills can support recurring work such as report formatting, code review, email drafting, or content development.

The distinction worth keeping straight, because it trips people up:

• Projects hold knowledge — the what Claude should know for a given piece of work.
• Skills hold procedure — the how Claude should carry out a specific kind of task, reusable anywhere.

If you want to create one, you do not have to hunt through menus. You can ask Claude to help you build it, then refine it until it matches the way you actually work.

Connectors: giving Claude access to your other apps

A Connector links Claude to an app or service you already use, such as Gmail, Google Drive, Google Calendar, Slack, Canva, and other tools. This lets Claude retrieve current information when a task requires it, instead of forcing you to copy and paste everything into the chat. A connector does not mean Claude is constantly syncing in the background. It reaches out when the conversation calls for that information.

For most people, the best starter set is simple: email, calendar, and file search. Those are the places where connectors save the most time because they reduce the amount of manual copying and context switching.

Plugins — one install, several capabilities bundled together

A Plugin packages several capabilities into one install for a specific workflow, such as sales, legal review, marketing, or data analysis. Instead of setting up each piece separately, a plugin brings the related tools together for a defined use case. Once installed, its commands are usually available from the chat command menu.

The best advice is to install only what you will actually use. A long list of unused plugins does not make the tool more powerful. It usually just makes the workspace harder to manage.

Claude for Excel and PowerPoint (and Word, and Outlook)

Claude can also appear inside Microsoft Office as a sidebar add-in for Excel, PowerPoint, Word, and Outlook. That means you can get help with a spreadsheet, presentation, document, or inbox without constantly switching windows. After installing it from Microsoft AppSource and signing in, Claude opens as a panel inside the Office app. This can make the workflow feel more connected, especially when the same thread needs to support work across email, documents, spreadsheets, and presentations.

Bringing it together: a decision path, not a checklist

None of this is meant to be adopted all at once. A reasonable order to actually build the habit:

1. Start in Chat. Turn on extended thinking and web search. Turn off training data sharing in Privacy settings if that matters to you.
2. Set up one Project for whatever you do most often — the recurring client, the weekly report, the ongoing research thread. Load in the documents it should already know.
3. Notice when you’re asking Claude to build something, not just tell you about it, and let Artifacts do that work visibly instead of describing it in text.
4. Add one Connector — email or a shared drive — the first time you catch yourself copy-pasting content into a chat that Claude could have retrieved itself.
5. Only then look at Skills, Plugins, Cowork, or Code, once you know specifically what repeated problem you’re trying to solve. Extensions are answers to problems you should meet first, not features to collect for their own sake.

The people who get the most value from Claude are usually not the ones who turn on every feature. They are the ones who match a few features to the actual shape of their work. Start with the mode and two or three tools that support what you already do every week. Let the rest wait until you have a clear reason to use it.