Styles, Preferences and Memory

Three features personalise Claude without touching the content of any single prompt. Styles decide how Claude writes. Preferences tell it who you are. Memory lets it remember what matters to you from one conversation to the next. Each is useful, each has controls, and the exam expects you to know what each does, where it is managed and what it does not do.

Styles

A Style is a reusable instruction about voice and formatting. Claude ships with preset Styles (a normal default, a concise option for short answers, an explanatory option that teaches as it goes, and a formal option for polished business prose; the exact set varies). You choose a Style from a selector near the message box, and it applies to the current conversation until you change it.

Custom Styles are where the value lies for business users. You can create one by describing the voice you want or, more effectively, by giving Claude samples of writing in that voice. Claude analyses the samples and produces a Style you can refine. Typical custom Styles: "Our customer-facing voice", "Board paper register", "Internal Slack, informal", "Regulatory submission". A Style governs how something is said, not what is said; facts, structure and task still come from your prompt, Project and instructions. Styles are personal to your account; to share a voice across a team, describe it in a shared Project's instructions or in a Skill and reference examples held as knowledge.

Preferences

Your profile settings hold the information Claude should always know: what to call you, your role or field, and a free-text preferences field. These are the "personal preferences" layer from the previous lesson. Keep them short and universal: language variant, default detail level, profession, formatting habits. Anything that varies by task belongs in a Project instead.

Memory

Memory lets Claude retain useful context across conversations: your projects, your preferences for how certain tasks are done, people and terms you refer to often. With Memory on, you do not need to re-explain that you lead the EMEA team or that your quarter ends in March. Key characteristics, which the exam tests directly:

• Plan-dependent. Memory is available on paid plans and has rolled out progressively; business plans give administrators controls over it.
• Visible and editable. You can see a summary of what Claude has remembered in settings and edit or delete it. You can also simply tell Claude to forget something, or to remember something specific.
• Scoped per Project. Memory formed inside a Project stays within that Project, so client contexts do not bleed into each other, and work inside a Project does not alter your general memory.
• Switchable. Memory can be turned off entirely. With it off, each conversation starts fresh except for your preferences and any Project instructions.
• Incognito chats. An incognito conversation does not appear in your history and neither draws on nor adds to Memory. Use it for one-off sensitive questions or to test how Claude behaves without your context.

Memory is distinct from two neighbours. Project knowledge is reference material you deliberately curate; Memory is context Claude builds from how you work. And the ability to search past conversations ("what did we decide about the vendor last month?") is a retrieval feature over your chat history, available when you ask, rather than persistent memory that shapes every reply.

Some surfaces also allow importing context from other assistants you have used, so you are not starting from zero; check the help centre for current options.

Getting the three to work together

A consultant sets personal preferences (British English, concise by default, "I am a management consultant"). She creates two custom Styles from samples: "Client deliverable" and "Internal note". Each client has a Project with its own instructions and knowledge, and Memory inside each Project accumulates that client's context. For a quick personal question unrelated to work, she opens an incognito chat. The result is that a new conversation in the "Northwind" Project, with the "Client deliverable" Style, already knows who she is, how she writes for clients, what Northwind's situation is and what was agreed last week, before she types a word.

Controls and hygiene

• Review your Memory summary every month or two; remove anything stale or wrong, because an incorrect memory produces confidently wrong context.
• Do not rely on Memory for facts that must be exact (contract values, dates); put those in Project knowledge.
• Use incognito for anything you would not want surfacing in a later conversation.
• On business plans, understand what your administrator has enabled before assuming Memory is on or off.
• When sharing a device or demonstrating Claude, remember that Memory may reveal context from earlier work.

Exam focus

• Styles govern voice and formatting per conversation; custom Styles are built from descriptions or writing samples and are personal to your account.
• Preferences are universal defaults about you; task-specific rules belong in Project instructions.
• Memory persists context across conversations, is viewable and editable, is scoped per Project, and can be switched off.
• Incognito chats neither use nor add to Memory and do not appear in history.
• Memory differs from Project knowledge (curated reference material) and from past-chat search (on-demand retrieval of history).