What do I need to set up?
The room, the folder and the habits a non-technical reader needs, using the Claude ecosystem as an example.
You do not need to be technical to start building with AI. You need the right room, a folder of your own, and three habits. This page is the room part; the habits are the primer.
Chat, Cowork, Code: what is the difference?
Three rooms, one Claude. The difference is what Claude can reach, and how much it does between your messages.
Chat is the conversation. You type, Claude answers; you paste things in, Claude reasons about them. It can search the web, write documents, and build small interactive pages you can see in the chat. What it cannot do is touch your files or your computer. Chat is where you learn, plan, draft and ask every question you are embarrassed to ask a colleague. Start here.
Cowork is a colleague at your desk. It lives in the Claude desktop app, you point it at a folder, and it works IN that folder: reading your documents, organising files, building spreadsheets and reports, running longer multi-step jobs while you do something else. No terminal, no code. If your project is “a folder of documents that needs work done to it”, Cowork is the room. (This very website began as a folder built in a Cowork session before a line of its code existed.)
Code is the workshop. Claude Code is the builder’s version: it lives in your project’s code, edits many files at once, runs the build, runs the checks, and shows you what changed. It is the room where software actually gets made, and the one where the veracity habits matter most, because the model can now do real work you must be able to verify. You can run it in a terminal, in an editor, or from the desktop and web apps; the terminal is only one door.
A rule of thumb for the non-technical reader: learn in Chat, work your documents in Cowork, build your product in Code. You will likely use all three in that order, and there is no shame in staying in the first two rooms for months.
What do I need to get started?
- A Claude account (claude.ai). The free tier is a real start, not a demo.
- The desktop app if you want Cowork, since that is where it works with your local files.
- A folder of your own for the project. One folder, clearly named, everything in it. The single most useful non-technical habit on this site is that the project IS the folder, and the model is always pointed at it, never at your whole machine.
- A place to keep your record. A notes file in that same folder is enough: what you asked for, what you approved, what you discarded and why, each with a date. The primer turns this into a discipline; on day one, a dated notes file is the whole of it.
Can I do this on a free account?
You can start, honestly. Checked 25 July 2026:
- Free gives you Chat on web, phone and desktop, with web search, file creation and the interactive pages, under a usage budget that resets on a cycle. Claude Code and Cowork are not on the free tier.
- Pro (20 US dollars a month, less if billed annually) adds Claude Code AND Cowork, more usage, and the wider model choice. This is the tier where vibe coding properly begins, because it is the tier where Claude can reach your folder.
- Max (from 100 US dollars a month) is Pro with a bigger budget, five or twenty times the usage, for people building daily. You will know if you need it, because the weekly limit will tell you.
The honest sequencing: start free in Chat, and learn to write a good brief there, because a good brief costs the same on every tier. Upgrade to Pro the week you have a real folder that needs real work. Consider Max only when the limits interrupt you more than once a week.
Where do the limits actually lie?
Every tier has a usage budget, and heavier models spend it faster. Three practical truths:
- Limits arrive as a pause, not a wall. Your budget replenishes on a cycle; plan sessions rather than marathons.
- The budget rewards short, well-aimed prompts. A vague prompt that needs five retries costs five times the clear one. The six-field brief in the primer is also a cost control.
- The newest, largest models are the scarcest. On paid tiers Claude offers a choice of models; use the strongest for judgement calls and design, the workhorse for routine execution. This is exactly how this site is built.
Numbers, names and tiers change often; this page states the prices once, keeps to shapes everywhere else, and carries a last-reviewed date for exactly that reason.
What should my first session actually be?
Open Chat, free tier, and bring one real thing: a document to improve, a plan to shape, a page to draft. Ask for it with a goal, a scope and a way to verify (“rewrite this letter, keep it under a page, keep every date and figure unchanged, list what you changed”). Read what comes back the way you would read a colleague’s work: gratefully, and checking every claim.
Here is one that costs you nothing. Say you have a page of rough notes from a meeting. Paste them in and ask: “turn these into three clear action points, keep every name and date exactly as written, and tell me anything that is ambiguous rather than guessing.” What comes back is usable in seconds, and your job is the veracity: check each action against the notes, and notice whether it flagged the uncertain parts or quietly filled them in. That one exchange is the whole discipline in miniature.
That is vibe coding with veracity in one message, and everything after it is scale.