How to properly set up Claude Cowork

The setup guide I wish I had on day one.

Hey friends 👋

Welcome to another Signal Pro workflow issue.

In last Sunday’s newsletter, a reader's question hit a nerve:

“I feel like I’m falling behind with AI. There’s so much happening and I don’t know where to start. What should I do?”

My answer was simple. Open Claude. Tell it everything about your job. Your responsibilities, your weekly tasks, the problems that eat up your time. Give it the full picture. Then ask it how AI can actually help you.

That advice resonated with a lot of you.

But here’s what I’ve noticed. Even the people who use Claude well, who’ve set up memory, who use projects, who’ve dialled in their prompts, are still doing one thing manually: the actual work.

Claude gives you a great answer in chat. Then you copy it. Paste it into a document. Reformat it. Save it. Open a spreadsheet. Paste something else. Build slides. Export. Rename. Upload.

The AI is doing the thinking. But you’re still doing all the doing.

There’s a tool that collapses that entire loop. And I think it’s the most underrated thing Anthropic has shipped.

It’s called Claude Cowork.

Instead of chatting with Claude and manually turning its outputs into work, Cowork works directly on your files. You point it at a folder, describe what you need, and walk away. It reads your documents, creates new ones, builds spreadsheets with working formulas, generates presentations, and saves everything directly to your computer. No copy-pasting. No reformatting. Just finished work.

Today I’ll show you:

• What Cowork actually is (and how it differs from Claude chat and projects)

• The 3 context files that give Cowork deep understanding of your role and preferences

• How to set up global instructions that apply to every session

• Real workflows you can steal and use today

• Scheduled tasks that run while you’re busy

What is Cowork?

Claude’s web app is excellent at conversation, memory, and projects. But it lives in a browser tab. Your actual work lives in folders, documents, and spreadsheets on your computer. Cowork bridges that gap.

It’s a feature inside the Claude Desktop app that moves Claude from conversation into execution mode.

You point it at a folder on your computer and describe what you need done. Claude reads your files, creates new ones, edits existing ones, and saves everything directly to your disk.

Here’s what makes it different from Claude chat (including projects):

Chat gives you answers. Cowork gives you files saved directly to your computer. Claude’s web app can generate Office formats too. But with Cowork, those files land straight in your folder ready to open, edit, and send. No downloading from a chat window and having to save to your local files manually. You describe what you need, and the finished deliverable appears where you actually work.

Chat responds to prompts. Cowork executes tasks. You describe an outcome (”create a formatted report from these meeting notes”), and Claude plans the steps, breaks them into subtasks, sometimes coordinates multiple parallel workstreams, and delivers finished output. You can watch it work or walk away and come back to finished files.

Chat is conversational. Cowork is autonomous. Including scheduled tasks that run automatically on a cadence you set. Monday morning briefings, weekly report generation, and end-of-day task summaries all run while you do other things.

It’s built on the same technology as Claude Code (Anthropic’s developer tool), just without requiring you to touch a terminal. Anthropic actually built Cowork using Claude Code itself, taking them ~10 days to do so.

The setup that changes everything

Here’s the thing that separates people who think Cowork is “just another AI feature” from people who use it to genuinely transform their workflow.

Cowork is powerful out of the box. But there’s a catch that you have to overcome. Unlike Claude’s web app, which builds up memory over time and has project-level context, Cowork sessions start fresh. Each task is a clean slate.

This is actually a feature, not a bug. It means Cowork doesn’t carry over mistakes or outdated assumptions. But it also means that without setup, you’ll spend the first few minutes of every session re-explaining who you are and what you need.

The fix takes about 30 minutes. Once.

The secret is context files. Small text files (in Markdown format, so just plain text with an .md extension) that live in your working folder. Cowork reads these automatically at the start of every session.

Think of them as an onboarding document for a new employee. Except this employee reads the document every single morning before starting work, and never forgets a word.

And here’s the part I find most compelling. Remember the advice from my Sunday newsletter? “Tell Claude everything about your job”? Context files are the structured, permanent version of that. You write it once, and it informs every single task Cowork runs in that folder. Unlike memory (which captures fragments over time), these files let you deliberately design exactly what Claude knows about you, your role, and how you want things done.

Here’s what you need:

File 1: about-me.md — who you are, what you do, what your current priorities are, and what matters most to you right now. Your role, your team, your industry, your responsibilities.

File 2: voice-and-style.md — how you like things written and formatted. Your tone preferences, formatting rules, words you’d never use, examples of writing you like. This is the difference between output that sounds like you versus output that sounds like generic AI.

File 3: working-rules.md — your preferences for how Claude should behave. Should it ask clarifying questions before starting? Always save files in a specific format? Avoid certain approaches? This is your operating manual.

These files compound over time. You tweak them as your role evolves, your projects change, or you discover things Claude keeps getting wrong. Each tweak makes every future session slightly better.

The exact context files (copy, paste, customise)

Here are the templates I use. Customise these to your role and you’ll immediately notice a difference in output quality.