What You're Getting
Two paths to a personal AI that knows your context:
| The Simple Setup | The Full System | |
|---|---|---|
| What | Claude Desktop + Cowork | OpenClaw (self-hosted AI agent) |
| Effort | 10 min, no coding | 15 min, terminal required |
| Capabilities | Reads your files, helps with projects | Shell access, browser control, multi-channel, 24/7 |
| Safety | Sandboxed - can only read your folder | Real system access - can execute commands |
Start with The Simple Setup. If you want more power later, The Full System is there when you're ready.
The Simple Setup
10 minutes • No coding • Safe
Step 1: Get Claude Desktop
- 1.Go to claude.ai/download
- 2.Download Claude for Mac (Windows coming soon)
- 3.Install and sign in
Step 2: Create Your Second Brain Folder
Create a folder on your computer:
/Users/[you]/Documents/Second Brain/Inside it, create this structure:
Second Brain/
├── CLAUDE.md ← Instructions for Claude (master navigation)
├── Inbox/ ← New items to sort later
├── Projects/ ← Active work with deadlines
├── Areas/ ← Ongoing responsibilities
├── Knowledge/ ← Frameworks, learnings, notes
├── Resources/ ← Templates, bookmarks, reference material
├── People/ ← Your network and contacts
├── Skills/ ← Reusable instruction sets for Claude
└── Archives/ ← Completed or inactive itemsStep 3: Create Your CLAUDE.md
Create a file called CLAUDE.md in your Second Brain folder.
Copy this template and fill in the [brackets]:
# My Second Brain
## Who I Am
I'm [YOUR NAME].
Role: [What you do - e.g., Founder, Designer, Student, Manager]
Communication: [Direct / Detailed / Casual]
Timezone: [Your timezone]
## Current Focus
1. [Your #1 priority this month]
2. [Your #2 priority]
3. [Your #3 priority]
## Active Projects
| Project | Status | Next Action |
|---------|--------|-------------|
| [Project 1] | In progress | [Next step] |
| [Project 2] | Planning | [Next step] |
## Where Things Live
Every folder has an `_INDEX.md` that explains what belongs there.
Read the relevant _INDEX.md before working in any folder.
| Folder | What's Inside |
|--------|---------------|
| `Inbox/` | New items to sort later |
| `Projects/` | Active work with deadlines |
| `Areas/` | Ongoing responsibilities (health, finances, etc.) |
| `Knowledge/` | Frameworks, learnings, reference notes |
| `Resources/` | Templates, bookmarks, saved material |
| `People/` | Contacts and relationships |
| `Skills/` | Reusable instruction sets (read Skills section below) |
| `Archives/` | Completed or inactive items |
Each project folder has a `_TODOS.md` with four sections:
Critical, This Week, Next 30 Days, Someday.
## How to Work With Me
### Session Start
1. Check what I ask for. Load only the relevant context
2. If resuming previous work, check the project's _TODOS.md first
3. Don't load everything at once. Read _INDEX.md files to navigate.
### During a Session
- If the topic shifts, check the relevant folder's _INDEX.md
- If you learn something worth saving, tell me and save it to Knowledge/
- Keep responses focused. I don't need long explanations for simple tasks.
### Session End
- Update the project's _TODOS.md if we completed or added tasks
- If we discovered something useful, save it to the right place
## Skills
Skills are markdown files in `Skills/` that give you specialized
instructions for recurring tasks. When I mention a task that matches
a skill, read the skill file first and follow its instructions.
| Skill File | When to Use |
|------------|-------------|
| [e.g., content-writer.md] | [When I ask you to write content] |
| [e.g., meeting-prep.md] | [When I have an upcoming meeting] |
| [Add your own as needed] | |
## Commands
| I Say | You Do |
|-------|--------|
| "What's my status?" | Read Projects/_TODOS.md files, summarize what's active |
| "Plan my day" | Check priorities, suggest what to focus on |
| "Save this: [idea]" | Create a file in the right folder |
| "Process inbox" | Read Inbox/, help me sort items to the right folders |
| "Weekly review" | Review all _TODOS.md files, flag overdue items, suggest next week's focus |
| "[Project name] status" | Read that project's _TODOS.md and _INDEX.md, summarize |
## Goals (This Quarter)
- [ ] [Goal 1]
- [ ] [Goal 2]
- [ ] [Goal 3]Want Claude to set this up for you?
Skip the manual setup. After connecting your folder in Step 4, paste the Bootstrap Prompt below and Claude will create everything automatically.
Quick start: The bootstrap prompt
Instead of creating all the folders and files yourself, paste this prompt into Claude Cowork after connecting your folder. Claude will set up your entire system and ask you a few questions to personalize it.
I want you to set up my Second Brain folder as a personal knowledge system.
Here's what I need you to do:
1. CREATE THIS FOLDER STRUCTURE:
- Inbox/
- Projects/
- Areas/
- Knowledge/
- Resources/
- People/
- Skills/
- Archives/
2. IN EACH FOLDER, create an _INDEX.md file that explains:
- What belongs in this folder
- How files should be named (use: YYYY-MM-DD-descriptive-name.md)
- Example entries
3. CREATE THESE STARTER FILES:
- Projects/_TODOS.md (with sections: Critical, This Week, Next 30 Days, Someday, all empty)
- Inbox/WELCOME.md (a short welcome note explaining how to use the inbox)
- Skills/example-skill.md (a template skill file showing the format)
4. CREATE THE CLAUDE.md FILE:
Before writing it, ask me these questions (one at a time):
a. What's your name and what do you do?
b. What are your top 3 priorities right now?
c. What active projects are you working on?
d. How do you prefer Claude to communicate? (direct/detailed/casual)
e. What recurring tasks do you want to turn into Skills?
Then generate a CLAUDE.md based on my answers, following this structure:
- Who I Am (identity and preferences)
- Current Focus (top 3 priorities)
- Active Projects (table with status and next action)
- Where Things Live (folder map)
- How to Work With Me (session start/during/end instructions)
- Skills (table of skill files)
- Commands (what I say → what you do)
- Goals This Quarter
5. AFTER EVERYTHING IS SET UP:
Give me a summary of what was created and suggest 3 things I should add first.
Important rules:
- Use sentence case for all text (not all lowercase, not Title Case)
- Don't delete anything that already exists in this folder
- Ask me before making assumptions about my preferencesThis creates the folder structure, _INDEX.md files for each folder, starter files, and a personalized CLAUDE.md based on your answers.
How the knowledge system works
The structure above isn't just organization. It's how Claude navigates your brain. Three key patterns:
CLAUDE.md is the map
Claude reads this first every session. It tells Claude who you are, where things live, and how to behave. The better your CLAUDE.md, the less you repeat yourself.
_INDEX.md files are entry points
Each folder has an _INDEX.md that explains what's inside. Claude reads these to navigate, so it loads only what's relevant instead of reading everything. This keeps it fast and focused.
_TODOS.md files track work
Each project gets its own _TODOS.md with four priority levels (Critical, This Week, Next 30 Days, Someday). When you say “what should I work on?”, Claude reads these and gives you a real answer.
Skills: Give Claude specialized abilities
Skills are markdown files that turn generic Claude into a specialist. Put them in your Skills/ folder and reference them in your CLAUDE.md.
Content writer
Your voice, tone, audience, and format preferences. Claude writes like you instead of like a robot.
Meeting prep
How to research attendees, generate talking points, and create agendas from your calendar.
Weekly review
Steps for reviewing your projects, updating priorities, and planning the next week.
Research assistant
How to gather info, organize findings, and save them to your Knowledge folder with sources.
Here's what a skill file looks like:
# Content Writer
## When to Use
When I ask you to write content, a post, an email, or anything public-facing.
## My Voice
- Tone: Conversational but informed. Like explaining to a smart friend.
- Length: Keep it concise. Cut filler words.
- Format: Short paragraphs, use subheadings for longer pieces.
## Process
1. Ask me: who is this for, and what should they do after reading it?
2. Draft an outline (3-5 main points)
3. Wait for my approval before writing the full draft
4. After writing: flag anything that sounds generic or salesy
## Avoid
- Corporate jargon, buzzwords
- "In today's world...", "It's important to note..."
- Passive voice unless quoting someoneThe bootstrap prompt creates a template skill file for you. Customize it, add more as you need them. The more specific your skills, the better Claude performs.
Step 4: Connect Claude to Your Folder
- 1.Open Claude Desktop
- 2.Click “Cowork” tab at the top
- 3.Click “Select folder”
- 4.Choose your Second Brain folder
Done!
Try saying: “What do you know about me?”
The Full System
15 minutes • For builders • More power, more responsibility
What is OpenClaw?
OpenClaw is a self-hosted AI agent platform. It runs on your own hardware and gives your AI real capabilities:
How It Works
OpenClaw runs entirely on your hardware with three components:
Connects to WhatsApp, Telegram, Slack, Discord, iMessage. Routes messages to your local agent.
Runs on your machine. Executes shell commands, controls browser, manages files, accesses APIs.
Your choice: Claude, GPT, or Gemini. API calls go directly from your machine. Nothing routes through OpenClaw servers.
What Makes It Different
Persistent Memory
Stores context as local files. Remembers conversations and actions from weeks ago. No cloud sync. Your data stays on your machine.
Proactive Outreach
Runs continuously. Can send you updates, reminders, and alerts before you ask. Not just reactive. It anticipates your needs.
Autonomous Execution
Doesn't just suggest steps. Executes them directly. Runs scripts, controls apps, manages files. Action, not advice.
What People Actually Build
Accesses your calendar, sends WhatsApp 30 min before meetings with attendee backgrounds, company news, and talking points.
Handles sponsorship and vendor emails. Qualifies leads, negotiates within preset limits, escalates when needed.
Daily scans of competitor websites, LinkedIn, Twitter, and Product Hunt. Alerts you to pricing changes, new features, hires.
Aggregates Zendesk tickets, App Store reviews, G2 feedback, and Twitter mentions into a single prioritized dashboard.
Scans Google Drive contracts, tracks renewal dates and obligations. Alerts you 30 days before deadlines so nothing slips through.
⚠️ Critical Security Warning
OpenClaw gives your AI real system access: shell commands, browser control, file management. This power comes with real risks:
- •Security researchers found 900+ publicly exposed OpenClaw instances on Shodan with zero authentication. Anyone could send commands.
- •Prompt injection attacks are possible. A malicious email could trigger the AI to forward your credentials.
- •The creator himself says: “Most non-techies should not install this. It's not finished.”
Only proceed if you're comfortable with terminal, networking, and security configuration. The Simple Setup is safer and works perfectly on its own.
What You'll Need
- Node.js v22+ - Check with
node --version(install from nodejs.org if needed) - Anthropic API key - Get one at console.anthropic.com (starts with
sk-ant-, $5 free credits) - A messaging app - WhatsApp, Telegram, or any of the 13+ supported channels
- 15 minutes and comfort with terminal
Step 1: Install OpenClaw
Quick install (recommended):
curl -fsSL https://openclaw.ai/install.sh | bashnpm install (if you prefer):
npm install -g openclaw@latestIf you get “command not found” after install, run: export PATH="$PATH:$(npm prefix -g)/bin"
Step 2: Run Onboard
Run the onboard wizard - it will guide you through configuration:
openclaw onboard --install-daemonThe wizard will prompt for your API key (get one at console.anthropic.com) and preferences like workspace path.
Step 3: Connect WhatsApp
This step requires you to manually scan a QR code - it cannot be automated:
openclaw channels login- 1.A QR code will appear in your terminal
- 2.Open WhatsApp on your phone → Settings → Linked Devices → Link a Device
- 3.Scan within 20 seconds (QR refreshes automatically - this is normal)
Step 4: Verify Setup
Check that everything is running:
openclaw doctorThen send yourself a test message on WhatsApp. If your AI responds, you're done!
Useful Commands
| Command | Purpose |
|---|---|
| openclaw status --all | Check if running |
| openclaw doctor | Diagnostic check |
| openclaw logs --follow | View errors |
| openclaw gateway --port 18789 | Start gateway manually |
| openclaw security audit | Check security config |
| openclaw channels status | Check channel connections |
| openclaw uninstall | Uninstall gateway + data |
Troubleshooting
| Issue | Solution |
|---|---|
| command not found: openclaw | export PATH="$PATH:$(npm prefix -g)/bin" |
| QR code keeps refreshing | Normal - scan within 20 seconds |
| AI not responding | Run openclaw doctor, then openclaw gateway --port 18789 |
| Check errors | openclaw logs --follow |
Security Features
- Pairing mode - Default DM policy ensures only approved contacts can communicate
- Allowlist mode - Restrict access to specific contacts only
- Security audit - Run
openclaw security audit --deepfor comprehensive checks, add--fixfor auto-remediation - Sandbox config - Configure isolation in
openclaw.json
Safety Recommendations
If you proceed with OpenClaw, follow these practices to minimize risk:
- 1.Use dedicated hardware: Run on a separate Mac Mini, VPS, or old laptop. Never on your primary work machine.
- 2.Never expose ports publicly: Use Tailscale for private networking. No port forwarding, no public IPs.
- 3.Enable Docker sandboxing: Add
sandbox.mode: "non-main"in your config to isolate command execution. - 4.Start with low-risk integrations: Connect a secondary email or test calendar first. Delay banking, primary email, or sensitive accounts.
- 5.Consider managed deployments: If security configuration feels overwhelming, use DigitalOcean or Cloudflare options below.
Managed Deployment Options
Want OpenClaw's power without managing security yourself? These options handle the infrastructure:
One-click marketplace deployment. Pre-configured security, automatic updates.
View on Marketplace~$5/month. No hardware required. Runs on Cloudflare's edge network with Zero Trust auth, R2 storage, and semantic search over your notes.
View complete setup guideThese options trade some customization for security and convenience. Good for testing before self-hosting.
Uninstall
If you ever need to remove OpenClaw:
openclaw uninstall
npm uninstall -g openclawPreview what will be removed first with: openclaw uninstall --dry-run
What You Can Do
With The Simple Setup
- Work with your files and projects
- Write and edit documents
- Process your inbox
- Deep work sessions at your desk
With The Full System
- Message your AI from anywhere
- Execute commands and scripts
- Automate browser tasks
- 24/7 availability, voice interaction
Want More AI Systems?
Join the newsletter for weekly guides on building personal AI systems.
Get it in your inbox
Free weekly newsletter. AI workflows, technical deep-dives, and lessons from building in public.
No spam. Unsubscribe anytime.
Frequently Asked Questions
Get More
Questions? DM me on Instagram or Twitter
Weekly insights: Be Superpowered Newsletter
Full AI setup guides: Blue Orchid