Chapter L1.3 — Settings and styles

Level 1 — Foundations. Product details verified on 22/06/2026 against official sources.

Goal

By the end you will know how to configure Claude so it answers in your tone and with the right capabilities, without repeating the same instructions in every chat. We’ll look at account preferences, Styles and the Web search, Code execution and Memory capabilities.

Prerequisites

• Knowing how to start a conversation (ch. L1.1).

Account-level instructions

In Settings → Instructions for Claude you set preferences that apply to all conversations: what your name is, what language to write in, the tone you prefer, recurring terms from your work. It’s the right place for the things you’d say in every chat: writing them once saves you time.

It’s different from an instruction given in a single chat, which applies only there: these apply everywhere. Useful examples to put here: “always write in English”, “I prefer concise answers with examples”, “I’m a developer”, “don’t use emoji”. Keep them short and update them when your needs change: it’s the most convenient level for stable preferences.

Response styles

Styles control how Claude communicates. There are a few presets — Normal, Concise, Formal, Explanatory — and you can create custom ones, even by uploading samples of your writing. You’ll find them in the “Search and tools” menu next to the text box.

When to use them? Concise for quick, dense answers; Explanatory when you’re learning and want more context; Formal for professional texts. The custom style is the most powerful: you upload some of your own texts and Claude imitates their register and rhythm. It’s the first step toward “making Claude sound like you,” which we go deeper into with Skills in Level 5.

Warning: Styles are evolving toward Skills (see Level 5). Features and names may change: it’s a volatile area, so check the current state in the settings.

The capabilities to know

Beyond tone, you can turn on capabilities that broaden what Claude can do. The table summarizes them; below, the detail.

Table L1.3.1 — Main capabilities and availability.

Web search. Turns on live search on the web, with source citations. It’s indispensable for recent information, where the model’s “from memory” knowledge isn’t enough. It’s activated from the “+” button or the tools menu. It consumes your usage limits. On Team and Enterprise it must be enabled by an administrator. In practice: ask “search for the latest news on X” and Claude brings results with the links, so you can trace the source. Without it, it answers with what it “knows” and on recent facts it can be wrong.

Code execution and file creation. Gives Claude a sandbox environment to run code and produce real files: Excel, PowerPoint, Word, PDF, charts. It’s also needed to upload XLSX files. It’s available on all plans, with an indicative limit of ~30 MB per file. It’s what makes requests like “create an Excel with these expenses and a pie chart” possible: Claude writes the file and hands it back to you ready to download.

Memory and chat search. Memory builds a summary of your history, so Claude can draw on past context; it’s available on all plans. Search across past chats, by contrast, is reserved for paid plans. Incognito conversations (ghost icon) are excluded from both.

Privacy and data control

Three things to keep in mind. Incognito chats don’t end up in your history or in Memory: use them when you don’t want to leave a trace. Memory is a summary, not a faithful copy: if a fact is important, repeat it in the chat instead of trusting it was remembered. Finally, on Team and Enterprise some capabilities are enabled by an administrator from Admin → Capabilities: if you don’t see them, it’s not your fault, ask whoever manages the organization.

Tip: you don’t have to turn everything on right away. Start from Web search and Code execution: they cover most everyday cases; you add the rest when needed.

Projects, in a nutshell

A Project is a workspace with its own knowledge base and dedicated instructions, shared across the chats of that project. It’s useful when you work for a long time on the same topic. Free users can create up to five; sharing is reserved for Team and Enterprise. We go deeper into them in ch. L3.2.

Preferences, Styles or Projects?

They overlap, but they serve different purposes. Instructions for Claude are personal preferences that apply everywhere. Styles change how Claude communicates and you turn them on as needed. A Project isolates context and instructions for a specific topic or client, without polluting your other chats. Practical rule: stable preferences → Instructions; how to write → Styles; a standalone job → Project.

A concrete example: to write this book you could keep a Project with the style rules and the finished chapters, so each new chat starts “up to speed” without your having to repeat the context.

Note: the settings seen here are personal. In an organization, some choices (security, data use for training) are defined by the administrator: your account inherits those rules.

In practice: configure your Claude

1. Open Settings → Instructions for Claude and write your name, language and tone.
2. Choose a style from the “Search and tools” menu (try Concise).
3. Turn on Web search and ask a question about a recent fact.
4. In Settings → Capabilities, verify that Code execution is active.
5. Check the Memory settings and decide whether to keep it on.
6. If you’re in an organization, check in Admin → Capabilities what is enabled on your account.

Common mistakes

• Repeating the same instructions in every chat. Put them once in “Instructions for Claude”.
• Forgetting Web search on recent facts. Without it, Claude answers “from memory” and can get dates or numbers wrong.
• Expecting an Excel file without Code execution. It must be turned on, otherwise no file generation.
• Trusting Memory for everything. It’s a synthesis, not a precise archive: for important facts, repeat them in the chat.
• Thinking it’s a bug when a feature is missing. On Team/Enterprise the administrator may have disabled it: check in Admin → Capabilities.
• Creating a Project for every question. Projects are for ongoing work; for an isolated request a regular chat is enough.

Summary

1. Instructions for Claude applies your preferences to every conversation.
2. Styles control tone; they’re evolving toward Skills.
3. Web search is for up-to-date facts and consumes usage.
4. Code execution enables file creation; Memory remembers the history.
5. Search across chats is paid only; incognito stays excluded.

Next step

You’ve completed the foundations. In Level 2 we move on to local installation: we start from ch. L2.1 — Installing Claude Desktop.