Claude Context Starter Kit
Enhance Claude's capabilities by providing it with persistent context for better interactions.
Claude Context Starter Kit
Give your AI a brain before you give it a task.
The Problem
Every time you open Claude, it starts with a blank slate. No memory of who you are, what you're working on, what you've already told it, or how you like to communicate. You re-explain. It guesses. You correct it. Repeat.
That's not the AI failing — that's just how it works. But you don't have to accept it.
The Fix
A small, plain-text file called CLAUDE.md lives in your workspace folder. Every time you open Cowork, Claude reads it automatically. It learns who you are, what matters, and where things stand — before you say a word.
That's the whole idea. One file. Permanent context. No re-explaining.
How It Works
There are two levels of context:
| Level | File | What it contains |
|---|---|---|
| Workspace | CLAUDE.md (root) | Who you are, your preferences, all active projects |
| Project | projects/[name]/CLAUDE.md | Everything specific to one project |
The workspace file is your permanent brain. It persists across every session and every project. Start here.
Project files are loaded on demand. Your workspace CLAUDE.md includes a pointer for each project — when you bring up that project, Claude reads the project file automatically. You never lose context mid-project.
Quick Start (5 minutes)
Not a developer? Click the green Code button at the top of this page, then click Download ZIP. Unzip the file — you'll find everything you need inside.
Not sure where to start? Ask Claude: "Can you walk me through setting up this workspace?" and it will guide you step by step.
Step 1 — Get the workspace template
Copy CLAUDE.md from this repo into the folder you use with Cowork. Open it and fill in the sections that matter most right now:
- About Me — your role, what you work on, how you communicate
- Current Projects — your top 3 active projects (with the
Context:pointer for each) - Communication Preferences — how you like Claude to respond
Don't try to fill in everything. A half-complete file beats a blank file every time.
Step 2 — Select your folder in Cowork
Open the Claude desktop app, open Cowork, and select the folder where you just saved CLAUDE.md. Cowork will read it on every new session automatically.
Step 3 — Test it
Start a new session. Ask Claude: "What do you know about me and what I'm working on?"
If it can tell you — you're set. That's the baseline. Now you can work without re-explaining from scratch.
Step 4 — Activate the skills (optional but recommended)
Copy the skills/ folder from this repo into your workspace folder alongside CLAUDE.md. Then add Standing Instructions #12 and #13 from skills/README.md to your workspace CLAUDE.md.
Once activated, you can say "new project" to get a guided project setup, and "catch me up" to recover cleanly after a context compact.
Going Deeper — Project Folders & Skills
For any project you work on repeatedly, create a project folder:
your-workspace/
├── CLAUDE.md ← Your permanent workspace brain
├── SAVE_WORKFLOW.md ← How to save context between sessions
├── projects/
│ └── [project-name]/
│ ├── CLAUDE.md ← All context for this project
│ ├── CHANGELOG.md ← Project history
│ └── documents/ ← Files Claude produces for this project
├── scratch/ ← One-off tasks (no setup needed)
└── z-archive/ ← Completed projects (move, don't delete)
Pick a template from the templates/ folder that matches your work, rename it CLAUDE.md, drop it in your project folder, and fill it in. Then add the pointer to your workspace CLAUDE.md:
### Project: [Your Project Name]
**Status**: Active
**Goal**: [One sentence]
**Context**: Read `projects/[project-name]/CLAUDE.md` for full project context.
**Next Step**: [What's next]
That's it. Claude will load the project context whenever you mention it.
Using the skills
The skills/ folder includes two conversational workflows built into Claude:
- New-project wizard: Say "new project" and Claude walks you through setting up a project folder conversationally — no manual file creation.
- Compact recovery: Say "catch me up" after a long session and Claude reads your context files to confirm current state before continuing.
See skills/README.md for setup instructions.
What's Inside
| File / Folder | What it is |
|---|---|
docs/user-guide.md | Narrative walkthrough — what week one actually looks like, end to end |
CLAUDE.md | Master workspace template — your permanent AI brain |
CLAUDE-example.md | Fully filled-in example — see what a working file looks like |
SAVE_WORKFLOW.md | Step-by-step guide for saving context between sessions |
COWORK_SETUP.md | Installation and setup guide — install the app, select your folder, connect your tools |
SETUP_GUIDE.md | First-time setup checklist and workspace health check — verify your configuration is complete |
templates/writing-communication.md | For emails, reports, recurring communications |
templates/project-tracking.md | For managing tasks, decisions, and blockers across a project |
templates/research-summarization.md | For research projects, source tracking, and findings |
skills/ | Behavioral scripts for Claude — copy to your workspace to activate |
projects/ | One subfolder per active project |
scratch/ | One-off tasks — no setup, no SESSION files needed |
z-archive/ | Completed projects — move here, never delete |
Works With
| Interface | How it works |
|---|---|
| Cowork (Claude desktop app) | Reads your selected folder automatically — this kit is built around that behavior |
| Claude Code (terminal / developer workflow) | Reads CLAUDE.md from your repo or working directory automatically |
| Claude.ai (web / mobile) | Paste your CLAUDE.md content at the start of a session to activate context |
Same file. Every interface. No reformatting, no duplication.
Further Reading
If you want more depth before you start, or want to understand the architecture behind the kit:
- How to Give Claude a Memory (In About 15 Minutes) — step-by-step setup walkthrough for non-technical users, with troubleshooting and a health check prompt
- Why I Treat My AI Context Like Infrastructure — the architecture behind the two-level model, why plain text beats everything else, and why the timing matters
- A Week in the Life — narrative walkthrough showing the whole system working end-to-end, from day one through month one
The Growth Pattern
| When | What you have | What Claude can do |
|---|---|---|
| Day 1 | Workspace CLAUDE.md filled in | Knows who you are, doesn't make you re-explain |
| Week 1 | One project folder set up | Remembers full project context across sessions |
| Month 1 | Multiple projects + CHANGELOG entries | Tracks history, picks up where you left off |
| Month 3 | Lessons Learned section populated | Gets better at your work as you use it |
Philosophy
- Context is infrastructure. The five minutes you spend on a
CLAUDE.mdsaves you hundreds of re-explanations. - Plain text lasts. No app, no account, no sync required. These are just files.
- Start minimal. A partial file beats a blank file. Fill in what you know, ignore the rest.
- Projects stay separate. Don't mix everything into one giant file. Workspace for the permanent stuff, project folders for the specifics.
- Save your work. Claude's memory resets every session. The
SAVE_WORKFLOW.mdshows you how to end each session so the next one starts strong.
Contributing
Found something that works even better? Open a PR. This kit is built from real usage patterns with real non-technical users in a Cowork environment. Practical improvements are welcome.
Built for business users who want Claude to actually know them — no terminal required.