Who This Is For
This is for knowledge workers, founders, and operators who spend their days re-explaining the same context to Claude before they can ask anything useful.
It is especially useful if you are:
Making the same kinds of decisions every week (what to prioritize, what to draft, what to cut) and want Claude to reason from your actual context, not best-practice averages
Frustrated that "saved prompts" age out fast because the underlying context keeps changing
Running multiple types of work (client work, personal projects, content, ops) and need Claude to know where the lines are
Tired of three apps, two Notion pages, and a folder of unread voice notes that never come together into anything you can act on
This setup makes the most sense when you spend at least 5 hours a week inside Claude and have at least one role, project, or workflow where the same context shows up over and over.
How the Workflow Runs
Project Setup → Context Layer → Instruction File → Skill Building → Weekly Loop
What You Need
Claude desktop app (Pro, Max, Team, or Enterprise plan, free plans do not include Cowork)
A folder on your computer to act as the project home (the default location is
~/Documents/Claude/Projects/)10 to 15 documents you currently reference most often in your work
Optional: Google Drive, Notion, or other connectors via Settings → Connectors
Optional: Claude Code (the CLI) if you prefer working from a terminal with a literal CLAUDE.md file in your project root
Time to Set Up
90 to 120 minutes for the first full setup, but the system works better if you split it into three sessions of 30 to 40 minutes spread across one week.
After that: 20 to 30 minutes every Monday to update priorities, log new decisions, refine your skill files, and run your weekly review.Step-by-Step Walkthrough
Step-by-Step Walkthrough
Step 1: Understand what this workflow should and should not do
Set expectations before you build anything.
This setup can help with:
Giving Claude persistent access to your real files, priorities, and decisions across every session inside the project
Replacing the "first paragraph of context" you re-type in every prompt
Encoding how you like work delivered (tone, format, structure) so Claude stops producing generic outputs
Running recurring routines (weekly reviews, briefings, drafts) with consistent quality
Letting Claude reason across your context to surface things you have forgotten
It does not cover:
Replacing the human judgment behind decisions, you still make the calls
Reading files you never put in the project folder or connect to it
Remembering what happened in chats from a different project, project memory is scoped to the project
Running scheduled tasks while your laptop is closed, the desktop app needs to be open for those to fire
Step 2: Set up a Cowork Project
Open the Claude desktop app. Click the Cowork tab in the left sidebar, then click Projects. In the top right, click + New project.
You will see three setup paths:
Start from scratch (creates a fresh folder, you add files later)
Import a project (pulls in an existing Chat project)
Use an existing folder (wraps a Cowork project around a folder already on your computer)
For a second brain, "Use an existing folder" is the cleanest path if you already have a folder of working material. If you do not, "Start from scratch" creates the folder at ~/Documents/Claude/Projects/[your-project-name]/.
Before naming the project, pick a scope. One role, one client, one workflow. Not "everything I do." A brain that covers everything retrieves badly because Claude cannot tell which slice of the mess is relevant.
Good scopes: "Client Work, Acme", "PAI Labs Writing", "Board and Investor Comms", "Personal Finance".
Note: Cowork is only available on paid plans (Pro, Max, Team, Enterprise). If you are on the free plan, this workflow will not work as written. Claude Code (the CLI tool) is the developer-focused alternative, where the project instructions live in a CLAUDE.md file at the project root, and skills live in ~/.claude/skills/<skill-name>/SKILL.md.
Step 3: Replace the empty project with context
The default project has nothing in it. Replace that with your actual material.
Inside your project folder, create five subfolders:
active(current work, drafts, in-progress documents)reference(source material, research, background docs)templates(formats you reuse, structures, frameworks)archive(closed projects, old versions)inbox(unprocessed notes, voice memo transcripts, quick captures)
Then update the project with your:
Top 10 to 15 most-referenced documents, dropped into the right subfolder (do not migrate everything, this is the trap that kills the system in week two)
Scope statement as one clear sentence (what this brain is for, what it is not for)
Current top 5 to 10 priorities, saved as
priorities.mdActive projects with status, next action, and any blockers, saved as
active-projects.mdDecisions log with significant calls you have made and the reasoning, saved as
decisions.mdStanding instructions on how you like work delivered, saved as
standing-instructions.mdTwo to three examples of finished work you are proud of, pasted in full into
/referenceAn explicit out-of-scope list (topics, folders, or projects this brain should never pull from)
The inbox folder gets special treatment. Drop anything there without organizing it. Voice note transcripts, clipped articles, half-finished ideas. Claude processes it in batch later, when you run your weekly review.
Want the full breakdown?
This is where you get real AI workflows, prompts, and systems you can use to automate your work. If you're serious about using tools like Claude to grow your business, this is for you.
Unlock full access today