Most people treat AI like a blank-slate conversation every time. You ask a question, Claude answers, the session closes and the context is gone. It's like briefing a new consultant from scratch on every call and paying for the same onboarding over and over.
There's a better way. If you're using Claude Code or Claude's Cowork mode, a file called CLAUDE.md acts as persistent memory across your sessions. You write it once, Claude reads it every time, and you stop starting from zero. You're working with someone who already knows the playbook.
What a CLAUDE.md file does
Think of the difference between a new intern walking in every morning and a chief of staff who's been with you for two years. The chief of staff doesn't need your decision-making process explained, or your writing style, or why you killed that idea last quarter. They already have the context.
Your CLAUDE.md holds that kind of institutional knowledge. It's a markdown document where you tell Claude who you are, what you're building, how you like to work and what you've already decided. When I work with Claude now, I skip the preamble and get straight to work.
In practice, it means I no longer open every session like this:
- "I'm a brand strategist with 25 years in luxury wine and spirits. I write in a direct, dry tone, I hate rhetorical-question stacking, and I need formatting I can share with clients. Also, here's the ten-question framework behind my Branding 101 series..." (me, explaining myself for the fourth time that week)
- I just open a session and start working, because Claude already knows all of that.
Saved context makes the work faster. Claude stops offering generic advice, you can be more specific about what you need, and the collaboration moves quicker.
Where the file lives and how to create it
CLAUDE.md can live in two places, depending on your setup.
Global (applies to all your projects): put CLAUDE.md in a hidden .claude folder in your home directory. On a Mac that's ~/.claude/CLAUDE.md, and on Windows it's the same folder inside your user profile.
Project-specific: if different projects need different instructions (one wants a data analyst, another a product builder), put a CLAUDE.md in the project folder. Claude reads both, so the global file holds the basics and the project file adds the specifics.
Most people I know run a hybrid: a global file with the fundamentals (who you are, how you write, what you care about) and project files that layer on current context (what you're building, what's been decided, what you're testing).
What to put in it
Keep it short and practical. Write it like operating instructions and leave your life story out. Here's the structure I use:
## Who I am
Brand strategist and advisor, 25 years building global consumer brands (Treasury Wine Estates, Diageo, Mark Anthony Group). I advise premium and luxury brands on positioning and where AI belongs.
## Writing style
Direct and dry, with humor. Sentence case. No rhetorical-question stacking, no em dashes, no "that's not X, it's Y." Vary sentence rhythm between short and long.
## Current project: Branding 101
Ten-question brand positioning framework, one question per video. Posting two per week. Brand essence is always the last question.
## Working style
Default to constructive critic. Tell me what's wrong before what's right, and only praise work that earns it.Each section is specific enough to be useful and short enough to stay current. Think of it as sticky notes for your AI collaborator.
What to leave out
Leave out sensitive financial information, API keys, passwords and personal data you wouldn't want sitting in a file that persists across sessions. CLAUDE.md is useful, but it isn't encrypted, so keep it operational.
Skip the filler, too. "I like good ideas" does nothing. "I want design recommendations that prioritize accessibility and test against WCAG 2.1 AA" does plenty, and that specificity is what makes the file work.
If it isn't in the file, assume Claude doesn't know it. Anything critical goes in.
How to maintain it
Your CLAUDE.md should change as your work does. When you finish a phase of work, update it. When you make a big decision, add it. When your preferences shift or a project ends, edit it.
I review mine every quarter and add a "last updated" note. Over time it becomes a record of what you decided and why, which is weirdly useful even outside AI. Writing things down forces clarity, and I've caught my own inconsistencies that way.
Update the file when:
- A project phase ends or strategy shifts
- You make a big decision you don't want to re-litigate
- You notice Claude repeatedly missing important context
- Your working style changes (new team size, new priorities, new constraints)
Why this matters
The real value of CLAUDE.md is bigger than the three minutes of explanation it saves. It turns Claude from a generic assistant into something closer to a collaborator who knows your work, one you can be blunt with and who won't waste your time on options you've already ruled out.
In marketing we talk about how much faster aligned teams move. When everyone knows the strategy and what success looks like, execution gets tighter. CLAUDE.md creates that alignment between you and the AI you work with.
If you're serious about using AI as a working tool, stop treating every session like a fresh conversation. Set up the file, keep it specific, and update it when things change.
