IE University · Artificial Intelligence
The Claude Handbook
From chat to system. A working guide to setting Claude up properly, using it where the work actually happens, and building processes that keep running when you step away.
This document is generated and maintained automatically, following instructions set by socratista.com, and revised every month against Anthropic's own documentation and other public sources. Automation of this kind can carry errors or fall behind a product that changes weekly: treat anything decisive — a price, a limit, a command — as worth confirming at the source before you act on it. The official references are listed in §4.6.
Set it up once
- Account and privacy
- Importing your context
- Plans and pricing
- Tokens and models
Build a system
- Chat and Cowork
- Projects and memory
- Artifacts
- Connectors, skills, plugins
Let it run
- Scheduled tasks
- Dispatch and the browser
- Claude Code
- Safety and cost discipline
Zero to hero in two hours
- Install and survival kit
- Generate your CLAUDE.md
- Connect email or Drive
- Create your first skill
Why this matters now
Since 2023, using an AI assistant well meant writing a better prompt. In 2026, that is no longer where the difference lies.
The paradigm changed
prompt → answers
“waow”
You write the perfect instruction, every single time. It forgets you tomorrow.
context → answers + does
“waow waow”
You organise your context once. Claude answers and executes — on your files, your applications, your computer.
2023 was the year of learning prompts. 2026 is the year of building systems — a flow that runs without you in front of it: documented once, repeatable for as long as you need it.
The unit of value has moved from the prompt to the system: a configured environment that already knows who you are, what you are working on, and which sources it may use. Someone who only asks questions is using a fraction of what is available; someone who has set up context, tools and repeatable processes is working at a different rate.
Three ideas before you start
The expensive mistake is not the one you expect. Cost is the first question almost everyone asks, and the least useful one to open with. Tokens deserve watching — §3.6 is about precisely that — but what you risk in a month of learning is small next to what it buys you. The asymmetry runs the other way: at the speed this is moving, the real exposure is not an experiment that fails, it is a year spent watching while your field moves without you. Set a deliberate budget, then spend it without flinching.
Count your flight hours. Pilots are not measured by the manuals they have read but by the hours they have logged. The same applies here, and it gives you a metric worth more than any certificate: how many hours did you actually spend working inside Claude Code this week? Not reading about it — inside it. Two honest hours a week compound faster than a weekend course.
Be patient: Rome was not built in a day. The first version will often be good. It will rarely be the best one, and the distance between those two is where the value sits. So plan for several passes: each round starts from a better place than the last, which is why the third reaches somewhere the first could not, and the fifth somewhere the third could not. The fastest way up that curve is to ask Claude to change hats — to go over the same piece as an editor, then as a sceptical client, then as whoever will have to implement it. Each hat catches what the one before it missed, and each is, in effect, a specialist on call: the expert read that used to mean finding the right person, booking their time and waiting a week is now one request away. If you are unsure which hats the work needs, ask Claude which ones would catch the most.
What Claude is, beyond a chat window
- Chat and Cowork — chosen from the main text box. Chat is for thinking; Cowork is for delegating complete tasks.
- Projects group conversations, instructions, files and memory around one context.
- Skills teach Claude processes and criteria it can reuse whenever they apply.
- Plugins bundle skills, connectors, tools and specialised subagents.
- Scheduled tasks repeat work automatically, without you present.
- Connectors let Claude read and act inside external applications, within the permissions you grant.
This handbook takes those pieces in order, from first login to full automation.
How to use this handbook
Parts One to Three are the reference: what each piece is, when it earns its place, and the habits that keep it safe and affordable. Part Four is the practice: a two-hour guided session that assumes the reference exists and does not re-explain it. In a classroom setting, Part Four is the session plan and the rest is the pre-reading; on your own, you can run Part Four first and come back to the reference when a concept needs depth.
Foundations
Everything in this part is done once. Fifteen minutes of setup determines the quality of every conversation that follows.
Account and privacy
Create your account at claude.ai, either with a Google account or with an email verification code. When asked which plan you want, start on the free one. You will know when you need more, and §1.5 explains when that is.
Settle privacy before your first prompt
By default, Anthropic may use your conversations to improve future models. If you intend to work with client information, contracts or internal strategy, turn this off before you begin.
From that point your conversations sit outside training. Institutional and enterprise accounts are governed by their own agreement — if you are working under one, check what already applies before changing anything.
Bring your context across
If you have spent months with another assistant, you do not have to start from nothing. Claude can import the context that ChatGPT, Gemini or Copilot already hold about you.
How it works. Claude generates a prompt. You run that prompt in the other assistant, copy the answer, and paste it back. The transfer takes about five minutes.
Step by step
- In Claude: Settings → Capabilities → Memory → Start import.
- Copy the prompt Claude gives you.
- Open a new conversation in the other assistant — ChatGPT, Gemini or Copilot — and paste it.
- Allow 30–60 seconds for it to produce the full summary, then copy the whole answer.
- Return to Claude and paste it where indicated.
If you also want the full conversation history
- ChatGPT — Settings → Data controls → Export data. A ZIP arrives by email.
- Gemini — go to takeout.google.com, select My Gemini Activity and download.
The ChatGPT export email is not always delivered on the first request. Repeat the request until it arrives.
Two files worth writing once
An assistant that has never met you spends the first minutes of every session being briefed. Two short files retire that ritual for good.
ABOUT-ME.md
Your professional context: role, organisation, responsibilities, the work that recurs, the audiences you write for, the constraints you operate under.
ANTI-AI-STYLE.md
The phrasing, structures and habits you want Claude to avoid when it writes as you. This is what keeps output from sounding generated.
Build them from the data you exported in §1.2. Paste that material into Claude and ask it, in your own words, to turn it into each of the two files. Then download them and upload them to your Claude folder and to every project that matters, so they are read in every conversation in that context.
Web, desktop and mobile
Six ways in, each for a different moment. The first three are Claude itself; the last three put it where you already work.
Web
claude.ai from any browser. Create an account and start. The right starting point for everyone.
Desktop
macOS and Windows, from claude.ai/download. This is where Chat and Cowork are selected, and where local resources become available: files and folders, installed applications, the browser, computer use, Claude Code and Dispatch.
Everything in Part 3 lives here.
Mobile
App Store and Google Play. Text or voice, image and file analysis, downloadable documents and connectors. Cowork on mobile is still in beta and does not appear on every account.
Dispatch (§3.2) lets you send work to your computer from the phone.
Chrome extension
Claude alongside any page you have open. It reads what is on screen and answers about it, and — with permission — acts inside the page. Chrome only; §3.3 covers it.
Excel, and Office alongside it
Claude in Excel puts the assistant inside the spreadsheet, working on the sheet you have open. Microsoft 365 also connects as an application (§2.6), which reaches your documents from the conversation.
Claude Code
The most direct channel of all: Claude working on the files and folders of your computer, from a command line. It sounds technical and Part Four shows it is not. See §3.4.
The web, without hesitation. Move to the desktop app when you reach Part 3 and want Cowork or scheduled tasks.
Plans and what each includes
Move up a plan when the limits start interrupting your work, not before.
| Plan | Price / month | What it adds | Who it is for | When to move up |
|---|---|---|---|---|
| Free | $0 | The assistant itself, everywhere: browser, phone and desktop app. It searches the web, remembers you, builds files, runs code, holds projects and plugs into your applications. | Trying it out. Enough to judge whether this belongs in your week. | The moment a limit stops you mid-task — or the first time you want Cowork or Claude Code. |
| Pro | $20 less on annual billing |
Where the handbook actually starts: Claude Code and Cowork unlock here, along with Design, Science and Research, unlimited projects, the full model range and Microsoft 365. | Anyone using this as a working tool rather than a curiosity. | When you find yourself rationing your own work to avoid the ceiling. |
| Max | $100 — 5× $200 — 20× |
Pro with the brakes off: five or twenty times the session capacity, longer answers, features before everyone else, and priority when the service is busy. | People who spend hours a day inside it — building, analysing, running agents. | When hitting the Pro ceiling stops being an event and becomes a routine. |
| Team | Standard $25 / mo ($20 annual) Premium $125 / mo ($100 annual) |
The same tools with an organisation around them: one place to manage seats, single sign-on, search across the team's work, corporate connectors, and conversations excluded from training by default. Premium seats carry five times the usage of standard ones. | A department or a working group — two people is enough to start. | When the constraint stops being capacity and becomes governance: who has access to what. |
| Enterprise | Custom via sales |
Everything Team offers plus what an audit will ask for: automated provisioning, activity logs, a compliance API, retention you define, spend limits, network restrictions and a HIPAA option. | Institutions and regulated sectors, deploying to hundreds of people. | When legal, security or procurement need to be in the conversation. |
Prices are the published United States rates, before tax, as listed by Anthropic's Help Center on 27 August 2026. Currency and final amount vary by country. Team requires a minimum of two members ($25 / month standard seat, $125 premium; $20 and $100 on annual billing). Enterprise is contracted directly with Anthropic. Check claude.com/pricing for your region before deciding.
Tokens and models
Models do not read words; they read tokens, which are slightly smaller. Two or three rules of thumb are enough to reason about cost:
- 1 token ≈ ¾ of an English word (somewhat less in Spanish, where words run longer)
- 1,000 tokens ≈ 750 words
- One A4 page of text ≈ 500–600 tokens
Every message you send and every answer you receive consumes tokens — input and output both. That is what your plan measures.
| Model | Speed | Capability | What to use it for | Context window |
|---|---|---|---|---|
| Fable 5 | Slowest | Highest |
|
1,000,000 tokens max output 128,000 |
| Opus 5 | Medium | Maximum |
|
1,000,000 tokens max output 128,000 |
| Sonnet 5 | High | Very high |
|
1,000,000 tokens max output 128,000 |
| Haiku 4.5 | Fastest | High |
|
200,000 tokens max output 64,000 |
Opus 5 is the sane default for demanding professional work. Drop to Sonnet 5 when you want the same job done faster and cheaper, and to Haiku 4.5 when volume and speed dominate. Reserve Fable 5 for the longest, hardest projects — the ones that run for days.
One more dial: how hard it thinks
Besides which model, you can choose how long it reasons before answering. With extended thinking on, Claude works through a problem step by step — and shows you the steps, which is useful in itself: you can see where an argument went wrong instead of just distrusting the conclusion. In Claude Code the same dial appears as the effort level (§3.4).
Deeper thinking costs more tokens and more time, so treat it like the model choice: off for lookups and routine drafting, on for multi-step reasoning, numbers you need to trust, and decisions with real trade-offs.
Building a system
The setup is done. This part is about the structures that make Claude useful more than once: where work lives, what it remembers, what it can reach, and how it produces something you can hand over.
Chat and Cowork
The main text box is where you choose how Claude works. Chat is for conversation: thinking, analysing, researching, drafting. Cowork is different — Claude leaves the conversation and acts directly on your computer.
With Cowork active, Claude can open applications, read files from disk, edit documents, move through folders and run chained tasks. You give an instruction in plain language and it carries it out on your system.
Turning it on
Cowork is available only in the desktop app, on a paid plan. Select it in the text box. The first time, it asks for permission to reach your system — you decide which folders and applications it may see.
What it does
- Creates and edits files on your disk
- Reads documents and extracts information
- Runs tasks across several applications in sequence
- Works locally with your file system without you opening anything
What it will not do
- Irreversible actions without confirming with you first
- Anything you have not authorised
- Financial or security decisions on its own
You decide how far it goes.
Projects: your work organised by context
Conversations are islands: what you established in one is unknown to the next. A project turns them into a mainland.
It is a container where everything about one subject lives together — the conversations, a standing set of instructions, the documents that matter, and a memory of its own. You assemble it once and it is there every time you return.
What a project holds
Instructions
How Claude should behave inside this project: tone, role, constraints.
Context
The documents it needs to know — a service proposal, a brand guide, a client's data. Read in every conversation.
Memory
The key ideas from your conversations in this project, stored as it goes and kept separate from everything else.
History
Every conversation in the project, saved and reachable.
Creating one
When to use one
One project per distinct context. Mixing a marketing client with a legal matter and a personal blog costs Claude the thread, and costs you the thread too. Separate them, organise them, and Claude works in each as though it had been with you for months.
Styles: writing in your voice, not its own
Projects govern what Claude knows. Styles govern how it writes.
A style is a saved set of writing conventions you can apply to a conversation — register, sentence length, how much structure, what to avoid. You can define one from samples of your own writing, so that what comes back reads like you rather than like a model.
Styles and the ANTI-AI-STYLE file from §1.3 do the same job from two directions: the file describes what to avoid, the style describes what to do. Used together they are the difference between text you can send and text you have to rewrite.
Memory: managing what Claude knows about you
Memory is less mysterious than it sounds: Claude keeps fragments from your conversations and brings them back when they fit what you are doing.
Two kinds of memory
- Conversation memory. Claude extracts what matters from your conversations — what you do, how you work, preferences you have mentioned — and accumulates it on its own.
- Project memory. Inside each project, Claude builds a separate memory of the key ideas in that context. It does not mix with the rest.
Turning it on and managing it
Make sure both switches are on. From the same place you can see everything Claude has stored about you, edit it or delete it. It is worth reviewing every so often — memory sometimes keeps things that are out of date, or that you would simply rather it did not hold.
It does not replace ABOUT-ME. Memory is passive and built by Claude; ABOUT-ME is written by you, deliberately, and is far more precise. Use both — ABOUT-ME for what it should always know, memory for what it picks up along the way.
Artifacts: from conversation to deliverable
Ask for something that is not prose — a document, a model, a deck, a script — and the answer does not arrive as a wall of text in the chat. It arrives as an artifact: a real file, ready to download and use.
What you can create
- Word (.docx) — reports, proposals, contracts, templates
- Excel (.xlsx) — calculators, tables, models with formulas
- PowerPoint (.pptx) — structured, designed presentations
- HTML — landing pages, emails, web components
- Code — scripts in Python, JavaScript or any other language
- Images and diagrams — flows, mind maps, schematics
The conversation and the artifact sit side by side. Ask for changes in the conversation and the artifact updates as you go. When it is right, you download it in one click.
Do not open by asking for the finished artifact. Settle the content in conversation first — structure, data, tone. Ask for the artifact once that is solid. Artifacts cost more, and you do not want to spend that on half-finished versions.
Saved artifacts live in the sidebar, organised by date.
When the deliverable is visual: Claude Design
Artifacts cover documents, spreadsheets and code. When what you need is a screen — an app interface, a landing page, a poster, a one-pager — that is Claude Design's territory. Describe the piece in plain language and it proposes a designed version: layout, hierarchy, navigation, reusable components that keep every screen consistent. You refine it the same way you asked for it — by saying what should change — or by editing elements directly on the canvas, and the result can move on to Canva or to a presentation.
What it is for: putting something tangible in front of people before anyone writes code or opens a design tool. A founder validating an idea, a team agreeing on a flow, a first version of a campaign piece. What it is not: a replacement for a designer on complex work — and the vaguer your brief, the more generic the screens, so the thinking you would put into a §2.2 project brief belongs here too.
Claude Design is available on paid plans (§1.5) and is still a research preview — expect it to change. Its usage is metered on its own weekly quota, separate from chat and Claude Code, so exploring it does not eat the allowance you work with.
Connectors: Claude inside your applications
Out of the box, Claude is sealed off: it sees the conversation and nothing else. Connectors open doors — one per application — so it can look inside your tools and act there.
MCP stands for Model Context Protocol. You do not need to know what that means — only that it is the standard that lets Claude connect to almost any application.
What you can connect
Google Drive, Gmail, Notion, Slack, Calendar, GitHub, Canva, HubSpot, Jira, and a list that grows every week. If you use an application, there is a reasonable chance a connector already exists.
What it can do once connected
Not only read — act. Search files in your Drive, read email, create calendar events, update tasks in Notion, send messages in Slack. All from the conversation, without changing window.
Find the application, select it, and follow the process. In most cases it only asks you to sign in to your account.
A connected application is an application Claude can reach. Before connecting work email or corporate tools, check with your team or whoever manages security. Turn connectors off when you are not using them — you can do this directly from the message bar.
Skills and plugins: teaching it to work the way you do
General competence is the easy part; Claude arrives with it. What no model arrives with is your way of doing things — the shape of your reports, the rhythm of your scripts, the sequence you run when a client signs. Skills and plugins are where that goes.
Skills
A skill is a set of instructions telling Claude how to carry out a specific task. Not what to do, but exactly how.
A skill for LinkedIn posts, for example, might specify a one-line hook to open, no bold anywhere, a closing question, and a direct register throughout. Every time it runs, Claude follows that protocol without being told again.
Plugins
A plugin is a set of skills already packaged for a use case. Instead of building everything from scratch, you install it and get several capabilities at once — project management, content, data analysis, customer support and more. Some are published by Anthropic, others by the community.
Plugins can also be imported from a marketplace, or built with Claude.
You write skills, tailored to you. You install plugins, and they work straight away. The two combine without conflict.
Plugins installed from a marketplace are not vetted by Anthropic, and can change after you install them. Check the source before you trust one.
Advanced practice
Everything so far happens when you ask. This part is about work that happens without you: on a schedule, from your phone, in the browser, and inside your codebase — plus the discipline that makes all of it safe to run.
Scheduled tasks: work that runs itself
Everything so far waits for you to ask. A scheduled task does not wait.
A scheduled task is an instruction you give once, and Claude runs it whenever the condition you set is met — a specific time, a frequency, an interval.
What that looks like in practice
- Every Monday at 08:00, review the projects folder and produce a status summary
- Every day at 07:00, gather sector news and prepare a briefing
- Every Friday, analyse the week's files and write a report
- Every hour, check a folder and process any new files it finds
Write the instruction in plain language, define when it runs, and save. Claude handles the rest.
The task runs without you watching, but Claude reaches only what you authorised in Cowork — no permission, no access. It also needs your computer to be on and active.
Why it matters. The repetitive work you do every week — reviewing, summarising, preparing, organising — can leave your list entirely. You are not delegating it to a person. You automate it once and it stops existing as a task of yours.
Dispatch: directing your desktop from your phone
Cowork and scheduled tasks run on your computer. Dispatch closes the loop: it lets you direct Claude from your phone while the computer stays where it is.
Something occurs to you while you are out — a task, a request, an idea you want developed. Instead of noting it down for later, you send it to Claude, and it runs on your desktop.
Setting it up
- Install the Claude app on your phone (App Store or Google Play).
- In the desktop app, go to Dispatch.
- A QR code appears.
- Scan it with your phone's camera.
- The two accounts link automatically.
- Enable notifications so you are told as tasks progress.
Used normally, the mobile app is a chat: Claude answers, and acts nowhere. With Dispatch configured, what you send no longer goes only to the chat — it reaches the Claude on your desktop, which receives it as an instruction and carries it out with access to your system. The phone directs; the computer acts.
What it is not. Dispatch is not a way to work from your phone. The result happens on the desktop, not on the handset. It is for delegating while you are moving, not for replacing the computer.
Claude in Chrome: the browser agent
Claude in Chrome is an extension that installs Claude directly in your browser. From there it can read what you have open and, if you allow it, act inside web pages.
Installing it
Go to the Chrome Web Store and search for Claude — check that it is the official one — or install it from the desktop app.
Then activate it from your profile: Settings → Capabilities → Claude in Chrome. It works only with Google Chrome; Safari, Firefox and Edge are not supported.
Two things it does, and they are not the same
As a reading assistant
- Reads the page you have open and answers questions about it
- A long article, an offer, a contract — analysed without you copying anything across
As an agent
- Enters web pages and acts within them
- Fills forms, navigates, interacts with interfaces
- Here it no longer reads — it does
Use it calmly. Test in low-stakes environments first, and do not enable it on platforms whose terms prohibit automation.
Claude Code
Claude CodeClaude Code works directly with code and files from your computer's terminal. It is not a chat — it is an agent that reads, modifies and runs things inside your real project.
It requires a paid plan (Pro, Max, Team or Enterprise). The free account does not include access.
Installation is not optional
Claude Code appears inside the desktop app once you are on a paid plan, but nothing works until you install it from the terminal first. It is the prerequisite in every case.
macOS
curl -fsSL https://claude.ai/install.sh | bash
Nothing else needs installing beforehand, and it updates itself.
Windows, from PowerShell
irm https://claude.ai/install.ps1 | iex
PowerShell ships with Windows 10 and 11 — press the Windows key, type PowerShell and press Enter. Installing Git for Windows (git-scm.com) first is recommended, not required: with it, Claude Code gets its full command toolkit; without it, it falls back to PowerShell and still works.
Check that it worked
claude --version
The first time you run claude, a browser opens so you can sign in with your Claude account. There are no further steps; /login switches accounts later if you need to.
Two ways to use it
- From the terminal. Go to your project folder, type
claude, and work in plain language. It asks before modifying anything. - From the desktop app. Once installed, it appears as its own section. You can follow changes as they happen without touching the terminal.
What the desktop app adds
Model selection
Switch between Haiku, Sonnet and Opus whenever you like. Not every task needs the same model. Cowork, by contrast, does not let you change model once a conversation has started.
Effort level
Low, medium or high. It sets how hard the model works on an answer: more effort, more quality, more tokens.
Knowledge folder
A folder on your computer Claude may reach — reference documents, skills, templates, guides. Anything you put there is available without attaching it each time.
Usage panel
How many tokens you are consuming, which model is consuming them, and what share of your plan you have spent.
VS Code. Install the official Anthropic extension from the VS Code marketplace to see visual diffs and approve or reject each change with a button.
Who it is for. It was built for people who work with code — but it has stopped being only that. If you have never touched a terminal, do not start here: Part Four takes you from zero to a working setup in two hours, no programming required.
Working safely
Seven rules that cover most of what goes wrong.
Claude asks before it acts, so read what it proposes. Never approve blind.
Start in the most cautious mode. Manual — ask before acting — first; give it more autonomy once you know how it behaves.
Use Git whenever you use Claude Code. Commit before asking for significant changes — without version control there is no safety net.
Turn connectors off when you are not using them. Every active connector is an open door; close it when you do not need it.
Check before connecting institutional tools. Do not connect corporate systems without speaking to whoever manages security.
Review what it has done. Agentic AI makes mistakes, so read the result before treating it as finished.
Install plugins only from sources you trust. A malicious plugin can reach your data and your connected tools.
Spending tokens well
Six habits that make a plan go considerably further.
Let one model carry about 90% of the work. Opus is the sensible default for professional use; keep Fable for what is genuinely hard or very long. Drop to Haiku when a task is simple and you want it fast.
Markdown files are the most efficient format. A .md file consumes far fewer tokens than a formatted PDF or Word document.
Add files to the project, not to the conversation. Project files are indexed once; attaching them in every chat spends tokens repeatedly.
Keep instruction files short. Every line counts as tokens, and if a file runs too long, parts of it get ignored.
Compact a conversation when it grows long. Summarising the history frees context before you ask for something new and complex.
Watch the usage panel. Know how much you have spent and what share of the plan is gone before you hit the limit.
Claude Code in two hours
Claude CodeA timed onboarding sprint, written for people who are not especially technical — if you have never opened a terminal, this is for you. Start from zero; finish with Claude Code installed, a working profile, your first connected application and your first skill. Every block ends with something you can verify, and nothing you install along the way is exotic: the tools have names you may not know, but they are free, standard, and Claude handles them for you.
The plan, and where to run it
Two hours, five blocks. Nothing here requires programming — the skill you need is knowing how to ask, and this sprint teaches you the asking.
Where to run it
Claude Code runs in several places — the four below plus the web at claude.ai/code, JetBrains IDEs and Slack. All of them run the same agent; the difference is what you see while it works. (§1.4 compared Claude's apps in general — this table is specifically about Claude Code.)
| Environment | Strong at | Watch out for | Choose it if |
|---|---|---|---|
| Desktop app | Fully visual; changes and diffs on screen; model and effort selectors; usage panel | Needs a paid plan; slightly slower than the raw terminal | You are starting today. This is the recommended door in. |
| Warp a friendlier terminal |
A replacement for your computer's command line that behaves more like an app: each command sits in its own block, and it suggests what to type next | You install it separately and create an account; it uses more memory than the plain terminal | The terminal intimidates you but you would rather learn it than avoid it. |
| VS Code a free code editor |
You watch every file change happen as it happens, and approve or reject each one with a button before it is saved | It is a programmer's tool: a lot of interface to meet on day one if you have never opened one | You already work in VS Code, or you want to see every change before it lands. |
| Native terminal | Fastest, lightest, nothing between you and the agent | Text only; you read instead of watching | You are comfortable in a shell, or intend to become so. |
| Mobile app iPhone and Android |
Sends instructions to the Claude Code running on your computer, so you can start work from anywhere | The work happens on the desktop, not the phone — it has to be on and set up (§3.2) | Ideas reach you away from your desk and you would rather not lose them. |
| Chrome extension inside the browser |
Reads the page you have open and answers about it; with permission, fills forms and navigates for you | It works on web pages, not on your files — it is a companion to Claude Code, not a substitute (§3.3) | Your work lives in web tools rather than in files on disk. |
The environment does not make you better at Claude Code. Knowing what to ask does. Pick the door that intimidates you least and move on — you can change later without losing anything.
Install and survival kit
Block 1 · 0:00 – 0:20Installation is covered step by step in §3.4 — macOS is one command, Windows is one command plus Git. Do that first, confirm with claude --version, then open your working folder and type claude.
That is normal, and this is the moment to get past it. Everything this sprint installs is free, standard and removable. Expect it to happen the first time you ask for something real — a spreadsheet, a report, a web page — because Claude assembles those with tools it may not find on your machine yet. The names that come up most:
- Python and Node.js — the two workbenches almost everything is built on. Between them they cover spreadsheets, documents, charts and web pages.
- Git — the undo button for your projects. It records every change so any of them can be reversed; §3.5 insists on it for good reason.
- Libraries, requested by name as they are needed — small add-ons for one job each: writing a PDF, building a slide deck, drawing a chart, reading a spreadsheet.
- A local server, if you build something that runs in a browser — a landing page, a small application — so you can open it on your own machine before anyone else sees it.
None of these are programs you will ever open yourself. They are Claude's toolbox, the same one running on millions of computers, and they uninstall as easily as they arrive. When it asks, read the one-line explanation it gives, say yes, and let it work.
These are the commands that cover the first month. Everything else can wait.
| Command | What it does |
|---|---|
| claude | Starts a session in the current folder — that folder is what Claude can see |
| /init | Claude inspects the folder and drafts a first CLAUDE.md for it |
| /help | Every available command, in place |
| Esc, Esc | Interrupts Claude mid-task — your brake pedal |
| Shift + Tab | Cycles permission modes: manual (ask first) → auto → plan |
| /clear | Wipes the conversation and starts clean; your files are untouched |
| /compact | Summarises a long conversation to free context before a big request |
| /resume | Reopens a previous session where you left it |
| /mcp | Lists connected servers and handles their sign-ins (§4.4) |
| /exit | Leaves. Pressing Ctrl+D twice works too. |
Three files worth knowing from day one
- CLAUDE.md — the rules and context Claude loads at the start of every session. §4.3 builds yours.
- ~/.claude/skills/ — where your skills live, one folder per skill. §4.5 creates your first.
- .claude/settings.json — per-project permissions: what Claude may run, edit or never touch.
In your very first session, Claude Code asks before every change. From then on, on Pro, Max and Team plans, sessions start in auto mode: a safety reviewer checks routine actions and Claude edits most files without asking you each time. For these first two hours, press Shift + Tab until you see plan — Claude proposes before it touches anything — and loosen the setting when you trust what it proposes.
Your profile: generate your CLAUDE.md
Block 2 · 0:20 – 0:50Section 00 made the argument: the value has moved from writing perfect instructions to organising your context once. CLAUDE.md is where that starts. It loads at the beginning of every session — your role, your rules, your tools — so you stop repeating yourself in message one.
For a code project, /init drafts one from the folder itself. For your professional profile — who you are, regardless of project — fill in the form below. The file updates as you type; nothing you enter leaves this page.
If you wrote the ABOUT-ME file in §1.3, you have already done the thinking: same idea, different home. ABOUT-ME feeds your chat projects; CLAUDE.md feeds Claude Code sessions. Reuse its content here rather than starting again — the two files should agree with each other.
Where to put it
- ~/.claude/CLAUDE.md — global: loads in every session, every folder. Your profile belongs here.
- project folder/CLAUDE.md — per project: rules for that context only. Both load together.
Keep CLAUDE.md under 200 lines. It is loaded whole, every session — past a certain length Claude starts skimming it, which defeats the purpose. If it grows, the overflow is usually a skill in disguise (§4.5).
Connect your first application
Block 3 · 0:50 – 1:20§2.6 explained what connectors are; this block is only about doing it. One thing matters here: a connector set up once serves both chat and Claude Code, so these thirty minutes pay off everywhere.
The five-step route
- In Claude, open Settings → Connectors → Browse connectors.
- Search for the application — start with Google Drive or Gmail.
- Select Connect and read what it asks permission for. Actually read it.
- Authorise with the application's own sign-in.
- Done. A connector set up on this machine also serves Claude Code —
/mcplists what a session can reach.
One caveat before you click, and it is the one from §2.6: if this is an institutional account, have that conversation with your security people first.
The email-and-Drive test
- Connect Gmail or Google Drive as above (Outlook works the same way).
- Ask: "Find the most recent document I received about [a real topic] and summarise it in five lines."
- Then push it: "List the three most recent files in my Drive that mention [a real client or project], with dates."
Done when — Claude names real files with real dates, and you have opened one of them to confirm the summary is faithful. If it is not, say so in the chat: the correction sticks for the session.
Your first skill
Block 4 · 1:20 – 1:50You now have two places to teach Claude things, and the boundary between them is one sentence: CLAUDE.md is who you are. Skills are what you know how to do.
Belongs in CLAUDE.md
- Applies to every conversation
- Identity, language, tone, standing rules
- Loaded whole, every session — hence the 200-line cap
Belongs in a skill
- A specific, recurring task with its own steps
- Loads only when the task comes up — costs nothing the rest of the time
- One folder per skill; as many as you like
A consultant keeps name, language and formatting rules in CLAUDE.md — and proposal generation, invoicing and report layout as three separate skills.
meeting-summary
Create the folder and file — or simply ask Claude Code to create it for you, which is the point of having it:
- Create
~/.claude/skills/meeting-summary/ - Inside it, a file named
SKILL.md:
---
name: meeting-summary
description: Turn a meeting transcript or raw notes
into a structured summary with action items
---
When given a transcript or meeting notes:
1. Summarise in five lines maximum, plain language.
2. List decisions taken, one per line.
3. List action items as: owner — task — deadline.
4. Flag anything left unresolved under "Open".
5. Ask nothing; if information is missing, write "not stated".
- Paste any meeting notes into a session and ask for a summary — the skill loads by itself when the task matches its description.
Done when — the output follows your five rules without your having repeated them. That difference — instructions that persist versus instructions retyped — is the whole argument of this handbook in one file.
Checkpoint, and what comes after
Block 5 · 1:50 – 2:00Two hours in, you should be able to tick four boxes: Claude Code answers claude --version; your CLAUDE.md sits in ~/.claude/ and a new session already knows who you are; Exercise 1 found a real document through a connector; Exercise 2's skill fires on any pasted notes. Anything unticked, redo that block — the order matters less than the verification.
When connectors are not enough
Connectors cover roughly 80% of professional cases. For the rest, there is a ladder — climb only as far as you need:
- Connectors — first stop, always. Visual setup, OAuth, no terminal.
- MCP servers —
claude mcp addfor internal or custom tools; the protocol lets Claude discover what a server offers by itself. - Command-line tools —
gh,stripe,verceland friends. If a CLI exists, Claude Code can drive it. - Direct APIs — full control, for things headed to production.
- Web scraping — last resort. Fragile, and check the site's terms before you touch it.
Most working setups settle on two or three MCP servers plus a couple of CLI tools. Combine; do not collect.
If your organisation runs on Slack
On Team and Enterprise plans, Claude can join Slack as a colleague: mention @Claude in a channel and delegate. Each channel's Claude keeps its own memory and permissions, and pinned channel instructions act as its CLAUDE.md. Two habits from day one: pin those instructions before the first request, and have your administrator set a spending cap when enabling it — it runs on the most capable model, and unbounded first weeks show up on invoices.
What's next after today: the expert tier
Four ideas from earlier in this handbook become your next four sessions. Each one is a door you have already seen — now you have the setup to walk through it.
Cowork
Claude leaves the chat and acts on your computer: opens applications, reads your disk, edits documents, chains tasks. Desktop app → text box → Cowork.
Dispatch
Send instructions from your phone; Claude executes them on your desktop at home. The idea strikes on the street — it is done when you arrive. Setup: Cowork → Dispatch → scan the QR.
Scheduled tasks
"Every Monday at 8:00, review my projects folder and generate a status summary." Set once, runs alone.
Claude in Chrome
An extension that reads what is on screen and — with your permission — acts inside web pages, filling forms and navigating for you.
Claude Design
Describe a screen — an interface, a landing page, a one-pager — and get a designed prototype you refine in words or directly on the canvas.
And when even that feels small: put your projects under Git so every change is reversible (§3.5 already insists), deploy what you build with a single command, add a database when things need to persist, and split large work across parallel subagents. The sprint you just did becomes work that runs without you.
Keep learning: Anthropic Academy
Anthropic runs a free training platform at academy.claude.com, with courses in English and no cost to enrol. Two are worth every student's time before or after this sprint: AI Fluency: Framework & Foundations (14 lessons, about 4 hours — how to collaborate with AI effectively, efficiently, ethically and safely) and AI Capabilities and Limitations (13 lessons, about 3.5 hours — what these models can and cannot do, and why). The Academy also carries product tracks for Claude, Cowork, Claude Code, Claude in Slack and the Claude Platform, which pick up where this handbook stops. Catalogue as listed on 27 August 2026.
Official references
The facts in this handbook — prices, model specifications, commands, installation — were verified against Anthropic's own documentation on 27 August 2026. These pages are also where to check when something here has aged:
- Claude Code documentationEN— installation, commands, permission modes (§3.4, §4.2)
- Model specificationsEN— context windows and output limits (§1.6)
- Anthropic Help CenterEN— current plans and pricing (§1.5)
- Anthropic AcademyEN— free official courses
What good looks like
Nothing in this handbook is difficult on its own. What separates a useful setup from an unremarkable one is whether the pieces have been put together at all.
Time
Configured properly, Claude stops being one more tool to open and becomes the place a piece of work starts.
Output
A Claude that holds your context, your voice and your objectives produces work that bears no resemblance to the generic version. The gap is not marginal.
System
The point is not to use Claude occasionally. It is to build a flow that keeps working on your behalf — whatever your field or your level.