Set Up Your Personal Preferences
After this page you'll have a Personal Preferences block pasted into Claude Settings that makes every future conversation smarter, faster, and more like you, without you ever re-explaining who you are.
Estimated time: 5 minutes. Compounds forever after that.
The Name For This Thing
Personal Preferences is the always-true block about you. Claude reads it at the start of every conversation, in every Cowork folder, on every device. It's your baseline. If something about you is true everywhere, it goes here. If it's only true in one workspace, it goes in that folder's CLAUDE.md instead.
Think of it like this: Personal Preferences is your driver's license. CLAUDE.md is the badge for one specific office.
The 5-Minute Setup
Open Claude Settings, find Personal Preferences
Claude desktop, top-right menu, Settings, Personal Preferences. It's a text box. That's it. Confirm you can see it before continuing.
Write the four blocks
Four short paragraphs. Don't try to write a novel. Each block answers one question.
Block 1: Who you are
Name, role, situation. Mine: "I'm a PM building Salesforce workflow tools in regulated industries." One sentence. Two if you need it.
Block 2: How you want to be talked to
Short or long? Bullet points or prose? Casual or formal? Specific words you ban? Mine bans em dashes, "leveraging," and "piggyback." Be ruthless. The vaguer this block is, the more you'll regret it.
Block 3: The tools you actually use
Your daily stack. Google Drive, Slack, Notion, whatever. When Claude knows your tools it stops suggesting things you'll never install.
Block 4: Behavior rules
The "always do this, never do that" stuff. Mine: "Read MEMORY.md at the start of every session. Stay in character. Don't write to memory unless I ask."
Paste, save, test
Paste your four blocks into the Personal Preferences text box. Save. Open a fresh conversation and ask "What do you know about me?" Claude should reflect what you just wrote back at you. If it doesn't, the paste didn't save. Re-paste.
One next action
Use Claude for a real task right now. Draft an email, write a Slack message, summarize something. Notice what it gets right and what still sounds wrong. Those are the rules you'll add to your preferences on your next pass.
Three Rules That Save You Pain Later
Start small, then add. Don't try to write the perfect preferences on day one. Use Claude for a week, notice what annoys you, then add rules to fix it. My preferences have been rewritten about six times. Each rewrite started with a real moment of friction.
Be specific about what you hate. "Don't be too formal" is weak. "Never use 'leverage' or 'circle back'" is strong. Specific bans change output. Vague vibes don't.
Separate personal from folder-level. Personal Preferences are global. CLAUDE.md is local. Don't mix them. If a rule should fire in every conversation, it belongs in Preferences. If it only matters for one workspace, it goes in that folder's CLAUDE.md.
Further Reading
Claude Starter Kit (GitHub) →
My full setup as a paste-in-order kit. Includes a Personal Preferences template you can copy and edit, plus a workspace CLAUDE.md template and the Research / Plan / Implement / Validate prompts.
Ruben Hassid: Claude (How to AI) →
The minute-by-minute teaching style I borrowed for this page. Clear, no fluff, and Ruben's analogies for prompts vs projects vs skills are the cleanest I've read.
Lenny's Newsletter →
Lenny Rachitsky. PM-flavored writing teacher. The reason I name every concept and back it with a real example.
Get Started with Cowork (Official) →
Anthropic's official setup guide. The source of truth.
How to Properly Set Up Claude Cowork →
Alex Banks' 3-file method. The most thorough walkthrough I've found. (The Signal)
Claude Cowork Setup Guide (2026 Edition) →
Covers context files, global instructions, plugins, and workflows. (The AI Corner)
Part of Josh's Cowork Guide.