← home setup · markdown

CLAUDE.md

The master context file that turns a generic AI assistant into one that actually knows you and your business. Drop it in your project root and Claude reads it on every session.

56 lines · plain text 
# CLAUDE.md

## Who I'm working with
Name: [Your name]
Where: [Town and country, e.g. North Wales, UK. Helps with context, weather, time zone, local references]
Languages: [English, Welsh, etc. Tells the assistant what languages you actually use]
Background: [One line on what you've done, e.g. "20 years in software, ran an agency, now consulting"]
What I do: [What you actually do day to day in plain English]

## Current priorities
1. [Top priority for this month]
2. [Second priority]
3. [Third priority]

## Active projects
| Name | What | Status | Folder |
|------|------|--------|--------|
| [Project 1] | [One line on what it is] | [Active / paused / planning] | [path/to/folder] |
| [Project 2] | [What it is] | [Status] | [folder] |

## Key people
| Who | Role |
|-----|------|
| [Name] | [Their relationship to you, e.g. "mentor", "client", "brother who runs the warehouse"] |

## Terms
| Term | Meaning |
|------|---------|
| [Acronym or shorthand] | [What it actually means] |

## How I work
- Direct and concise. No over-explaining. Match my energy.
- Chat first. Only create a file if the task genuinely needs one.
- Plain English only. No jargon, no em dashes, no AI-sounding phrases, no corporate language.
- Strong drafts (90% there), not rough outlines.
- Tables over bullet lists for structured data.
- If a decision is needed, give 2-3 options with a clear recommendation.
- If I give clear instructions, get on with it. Don't check in.
- For big or irreversible tasks, show a brief plan first.
- Research first, then answer. Never ask me to do legwork you can do yourself.

## File rules
- Latest version only in the project root. Previous versions go to `archive/`.
- Each project folder has a `CONTEXT.md`. Read it when starting project work.
- Keep things tidy. Delete junk files immediately.

## Workspace map
```
[your-folder]/
├── CLAUDE.md         this file
├── TASKS.md          active tasks
├── projects/         [each project has its own CONTEXT.md]
├── context/          [working-style, brand-voice, design-principles]
└── memory/           [anything you want the assistant to remember]
```

Why this works

This is the single most useful file in my setup. It’s the brief I give every AI assistant before it does anything for me, on every project.

Without it, Claude has to guess. Guessing is where the generic, hedged, soulless responses come from. Three or four lines of real context, and you get a different assistant.

What this file does:

  • Tells the assistant who you are and what you actually do
  • Tells it how to talk to you (tone, format, length)
  • Gives it the picture of your business: priorities, projects, people, jargon
  • Sets the rules for files and folders so it doesn’t make a mess

How to use it:

  1. Copy the template
  2. Fill in the bracketed bits with your real details
  3. Save as CLAUDE.md in the root folder you work from (Claude Code reads it automatically)
  4. If you’re using the Claude desktop app, drop the same content into your project’s “instructions” or “memory”
  5. Update it weekly as priorities change

The rule: treat it like an onboarding document for a new hire. The more honest you are about how you actually work, the more useful the assistant becomes.

Most people never set this up. The ones that do see the difference instantly.

← back home