How I Turned My Obsidian Vault Into a Full Business Operating System Using Claude Code

Six months ago my Obsidian vault was a graveyard.

3,000 notes. Two years of captures. A folder structure I had reorganized four times and still could not navigate in under 30 seconds.

Every productivity YouTuber told me the second brain would change my life.

What they did not tell me was that a second brain without a nervous system is just a very expensive filing cabinet.

The filing cabinet changed the day I connected Claude Code to it via MCP.

Not connected as in “I paste things in.” Connected as in Claude Code has direct read and write access to every file in my vault in real time and operates as the intelligence layer on top of everything I have ever written.

Today the vault is not a note-taking app.

It is a full business operating system.

Research runs automatically. Content drafts itself. Projects update their own status. Weekly reviews write themselves. Client briefs generate before every call. The morning briefing is waiting when I open my laptop.

Here is the complete course for building the same system.

Why Obsidian and Not Notion, Roam, or Anything Else

Before the architecture, the tool choice matters.

Every other note-taking tool sits behind an API or a proprietary format.

Notion has a great API. But your notes are in Notion’s cloud, formatted in Notion’s proprietary blocks, dependent on Notion’s continued existence.

Obsidian is plain text Markdown files on your local machine.

When Claude Code connects to Obsidian via MCP it is reading and writing real files. Not making API calls to someone else’s server. Not converting proprietary blocks. Reading actual .md files the same way a text editor does.

This matters for three reasons.

First, speed. File system access is faster than any API. Claude processes vault content significantly faster than it processes API-fetched content.

Second, reliability. No API rate limits. No server downtime. No authentication failures. The files are there or they are not.

Third, ownership. Your vault is yours permanently. No subscription required to access your own notes. No migration required if the company changes its pricing model.

The business operating system I am going to show you depends on all three of these properties being reliable.

The Architecture: Five Systems That Run the Business

The Obsidian business operating system is not one big system.

It is five specialized systems that each own a specific business function and communicate through shared vault structure.

System 1: Intelligence and Research System 2: Content Production System 3: Client and Project Operations System 4: Personal Performance System 5: Financial Tracking

Each system has its own folder in the vault, its own CLAUDE.md configuration, its own skill files, and its own automation workflows.

Together they run every operational function of a solo creator or founder’s business.

The Vault Structure That Makes Everything Work

Before you build any system you need a vault structure designed for machine operations, not just human navigation.

Most Obsidian vault structures are designed for humans browsing. Folders organized by topic, by project, by date.

The operating system vault is designed for agents operating. Every folder has a predictable purpose. Every filename follows a consistent convention. Every note has a properties header that Claude can parse reliably.

00 - INBOX/ [unprocessed captures land here]

01 - PROJECTS/ [client-name]/ 00-overview.md 01-brief.md 02-notes/ 03-deliverables/ 04-communications/ [internal-project-name]/ same structure

02 - AREAS/ content/ finances/ relationships/ health/ learning/

03 - RESOURCES/ research/ references/ templates/

04 - SYSTEM/ CLAUDE.md skills/ workflows/ logs/

05 - DAILY/ [YYYY-MM-DD].md

06 - GENERATED/ briefings/ drafts/ reports/ syntheses/

07 - QUEUE/ [task files Claude processes automatically]

08 - ARCHIVE/ [completed projects and outdated content]

The GENERATED and QUEUE folders are the operational core.

QUEUE is how you assign tasks to the system without initiating a Claude session. You drop a file in Queue describing what you need. The automation layer picks it up, processes it, and deposits the output in GENERATED.

GENERATED is where every autonomous output lands. Never edit files here manually. They are Claude’s outputs. When you want to use something from GENERATED you copy it or reference it.

The Master CLAUDE.md

The CLAUDE.md file in your 04 - SYSTEM folder is the constitution of your business operating system.

Every agent, every workflow, every automated process reads this file before doing anything.

It tells Claude everything it needs to know about your business, your standards, your current priorities, and the rules it must follow when operating autonomously.

Business Operating System — CLAUDE.md

Owner

[Your name, your business, what you do, who you serve]

Business Context

Type: [Solo creator / founder / consultant / etc.] Revenue model: [How the business makes money] Team: [Solo / contractors / describe your structure] Stage: [Early / growth / established]

Vault Structure

[Brief description of what each folder contains]

Active Clients

[List each active client with engagement type and status]

Active Projects

[List each active project with status and next milestone]

Content Operation

Publishing schedule: [Your posting frequency and platforms] Voice: [How you write. Key patterns. What you never say.] Current content pillars: [Your main topic areas]

Revenue Targets

Monthly target: [Your number] Current MRR: [Update monthly] Key metric: [The number you watch most closely]

Operating Rules for Autonomous Actions

• Never delete any file without explicit confirmation
• Never send any communication without my review
• Never commit to deliverables or timelines on my behalf
• Always date-stamp generated files as YYYY-MM-DD
• When uncertain: deposit in GENERATED and flag for review
• Log every write operation in 04 - SYSTEM/logs/operations.md

Quality Standards

[What good output looks like in every category]

Current Weekly Focus

[Update every Monday — what matters most this week]

The Current Weekly Focus section is the most important part to keep updated. Update it every Monday morning. It weights every Claude decision toward what actually matters this week rather than operating from stale priorities.

System 1: Intelligence and Research

The research system is the one that saved the most time and the one that gets more valuable every month.

Before this system, research was the bottleneck on every piece of content I created. An hour of reading, synthesizing, and note-taking before I could write a single word.

Now research happens autonomously. I wake up to structured research briefs on topics I queued the night before.

The Research Skill File

Save this as 04 - SYSTEM/skills/deep-research.md:

deep-research

Trigger

“Run deep-research on [TOPIC]” or drop a file in Queue named RESEARCH-[topic].md

Context Required

Read CLAUDE.md for current business context and content pillars. Check 03 - RESOURCES/research/ for any existing notes on this topic.

Process

1. Identify the core question or angle most relevant to my audience
2. Synthesize what I already have in the vault on this topic
3. Identify the insight most people miss
4. Find the counterintuitive angle that creates interest
5. Locate 3 specific examples, statistics, or stories
6. Identify 3 content angles ranked by potential

Output Format

Save to: 06 - GENERATED/[DATE]-research-[topic].md

CORE INSIGHT: [one clear sentence] WHAT MOST PEOPLE MISS: [the non-obvious angle] SUPPORTING EVIDENCE: [3 specific examples with context] KEY DATA: [2-3 specific numbers or quotes] CONTENT ANGLES: [3 ranked angles for content creation] MY EXISTING NOTES: [relevant content from vault]

Quality Standard

If the core insight is something most people already know, it fails. Dig deeper until the insight is genuinely non-obvious.

The Automated Research Queue

N8N workflow that runs every morning at 5AM:

• Reads every file in 07 - QUEUE with names starting with RESEARCH-

• For each file, calls Claude API with the deep-research skill prompt

• Deposits the research brief in 06 - GENERATED/briefings/

• Archives the processed queue file

• Sends a Telegram notification with the count of research briefs ready

By 6AM every research task you queued the previous day is complete and waiting.

System 2: Content Production

Content is the output of the business. The content production system is the one most directly connected to revenue.

The system handles four content workflows automatically.

The Morning Content Brief

Every morning at 6AM an N8N workflow generates a content brief for the day.

It reads:

• Your most recent research briefs from GENERATED/briefings/

• The past 7 days of daily notes to understand what you have been thinking about

• Your current content pillars from CLAUDE.md

• What you have published recently (tracked in 02 - AREAS/content/)

Claude synthesizes all of it into a content brief with three angles ranked by potential, including the hook for each, the format that fits best, and whether it requires additional research.

You open your laptop at 7AM. The brief is in GENERATED. You pick one angle and start writing or you queue it for the content agent to draft.

The Content Draft Agent

Drop a file in Queue named DRAFT-[content type]-[topic].md with any specific notes you want to include.

Claude reads your research brief on the topic, your voice profile from CLAUDE.md, and the format guidelines for that content type. It produces a first draft in GENERATED/drafts/.

The draft follows your exact voice patterns because your CLAUDE.md contains a detailed voice profile extracted from your best-performing content.

My first drafts now require under 10 minutes of editing for 75% of pieces.

The Performance Analyzer

Every Monday at 7AM this workflow pulls your content performance data, feeds it to Claude, and deposits an analysis in GENERATED/reports/.

The analysis answers three questions every creator needs answered weekly:

What performed best and what specifically made it work.

What underperformed and the most honest reason why.

What to make more of and what to stop making.

The analysis feeds directly into the following week’s content brief, creating a closed loop where the system learns from its own outputs.

The Repurposing Engine

When a piece of content performs above a threshold you define, the repurposing engine automatically generates variants for other platforms.

A long-form article becomes 5 tweet variations, a LinkedIn post, and a newsletter section.

Each variant is format-specific and voice-matched. Not just reformatted but genuinely rewritten for how each platform’s audience reads content.

System 3: Client and Project Operations

This is the system that makes the biggest difference in the day-to-day feel of running a client business.

The Pre-Call Brief

30 minutes before any client call, the pre-call brief workflow fires.

It reads the client folder in 01 - PROJECTS, the last 3 meeting notes, any open tasks related to the client, the communications history, and the engagement overview.

It produces a one-page brief in GENERATED with:

• Relationship status and last significant interaction

• Open deliverables and their current state

• What was promised in the last call

• Suggested agenda items based on project status

• One thing to make sure you say before the call ends

I have not walked into a client call underprepared since I built this.

The Project Status Updater

Every time a file in a client’s project folder is modified, a workflow fires.

It reads the change, reads the existing project overview, and updates the overview with a timestamped status note reflecting what changed.

Your project overviews stay current automatically.

The Deliverable Tracker

A simple database maintained in Airtable (connected via MCP) that Claude updates whenever you log a deliverable completion in your daily note.

You write “DELIVERED: [client] — [deliverable]” in your daily note. Claude reads it, finds the client record in Airtable, marks the deliverable complete, and notes the date.

Your client delivery tracking stays accurate without a separate administrative step.

The Invoice Reminder

A workflow that runs on the 25th of every month.

It reads the client engagement overviews in CLAUDE.md to identify monthly retainer clients. It checks the Airtable payment records. It generates draft invoice reminder emails for any retainer clients who have not yet paid for the upcoming month.

The drafts land in GENERATED/communications/ for your review before sending.

System 4: Personal Performance

The operating system is not just for business functions.

The personal performance system tracks the inputs that produce business outputs: energy, focus, decisions, and habits.

The Daily Note Template

Every daily note generates automatically at 11PM for the following day.

It pulls from:

• Your content brief for tomorrow

• Any client calls on your calendar (via Google Calendar MCP)

• Open tasks from yesterday’s daily note that were not completed

• The current weekly focus from CLAUDE.md

You wake up to a structured daily note with your agenda, your open loops, and your focus for the day pre-populated.

The Weekly Review

Every Sunday at 8PM the weekly review workflow fires.

It reads all 7 daily notes from the past week, all project status updates, all content performance data, and all generated outputs from the week.

It produces a weekly synthesis with:

• What moved forward and the specific cause

• What did not move and the honest reason

• One pattern that appeared across the week worth naming

• Three priorities for next week ranked by leverage

• One personal insight from the week’s daily notes

The weekly review used to take 45 minutes of manual note-reading and reflection.

Now it takes 10 minutes to read what the system synthesized and 5 minutes to add anything the system missed.

The Decision Journal

Every time you write “DECISION:” in a daily note, an automated workflow captures the entry.

It asks Claude to add three fields to the decision record: the key assumption the decision rests on, the leading indicator that will tell you if the decision was right, and a date to check back.

90 days later the decision journal workflow surfaces those decisions for review.

You close the loop on every major decision you make.

System 5: Financial Tracking

The financial system is the most straightforward but consistently the most neglected.

The Revenue Tracker

A simple workflow that fires whenever you log a payment received in your daily note.

Format: “RECEIVED: $[amount] from [client] for [description]”

Claude reads the entry, updates the monthly revenue total in 02 - AREAS/finances/revenue.md, and updates the client payment record in Airtable.

Your revenue tracker stays current with zero additional administrative work.

The Monthly Financial Brief

On the first of every month, a workflow generates a financial brief for the previous month.

It reads all revenue logged in daily notes, all expense records, and the monthly target from CLAUDE.md.

It produces a one-page financial brief with:

• Revenue vs target

• Month over month comparison

• Client revenue breakdown

• Expense summary

• Projected vs actual for the quarter to date

You always know exactly where the business stands financially. Not at tax time. Every month.

Building the Automation Layer

The six systems described above require an automation layer that runs them without you initiating anything.

N8N is the right tool for this. Self-hosted on a $5 per month DigitalOcean droplet. No per-execution pricing. No usage limits.

The five workflow types you need:

Cron Workflows: Run on a schedule. Morning briefing at 6AM. Weekly review Sunday at 8PM. Monthly financial brief on the 1st.

File Watch Workflows: Trigger when files are created or modified in specific vault folders. Queue Processor watches the Queue folder. Project Updater watches project folders.

Event Workflows: Trigger from external events. Calendar events trigger the pre-call brief. Stripe webhooks trigger revenue logging.

Aggregation Workflows: Collect data from multiple sources and compile reports. Performance Analyzer pulls from multiple platforms. Financial Brief aggregates multiple data sources.

Notification Workflows: Send you updates when workflows complete. Telegram messages when briefs are ready. Email when something needs your review.

Start with the Cron workflows because they are the easiest to build and the ones that deliver the most immediate daily value.

The Queue Processor is the second priority because it is the most flexible. Once it is running you can assign any task to the vault at any time.

Add the Event and Aggregation workflows as you need them.

The Daily Experience After 30 Days

Here is what a typical weekday looks like once the system is running.

6:00 AM: Your phone receives a Telegram notification. Three research briefs and a content draft are ready in the vault.

6:05 AM: You read the morning briefing Claude generated. Your top three priorities for today. Two open loops from yesterday. One decision to make before noon.

6:15 AM: You review the content draft Claude produced overnight from the topic you queued yesterday. You spend 8 minutes editing. You schedule it.

8:00 AM: You open your project folder for your 9AM client call. The pre-call brief is already there. You read it in 5 minutes. You are more prepared than you would have been after 30 minutes of manual prep.

During the day: You capture ideas, tasks, and notes in your daily note using the simple conventions the system understands. Everything gets processed automatically.

5:00 PM: You write a brief daily review in your daily note. Three things that moved. One thing that got stuck. One decision you made.

That is it. The system handles the rest.

The Compounding That Happens After 90 Days

The most underestimated aspect of the Obsidian business operating system is not what it does in week one.

It is what it does in month three.

Every research brief adds to the knowledge base Claude draws from for future research.

Every content performance data point makes the content brief more accurate.

Every client interaction that gets logged makes the pre-call brief richer.

Every decision journal entry builds a record that surfaces patterns in how you decide.

The system does not just maintain the business.

It learns how the business works and gets better at serving it.

By month three the morning briefing is calibrated to your actual patterns.

By month six the research briefs are surfacing connections across months of accumulated vault knowledge.

By month twelve you have a business operating system that knows your business as well as you do.

And runs it better than you would alone.

Build the first system this weekend.

Start with the Morning Briefing and the Content Draft Agent.

Add one system per week.

By month two you have an operating system running your business.

By month six you have a business that compounds.

Follow @cyrilXBT for the exact N8N workflows, CLAUDE.md templates, and skill files that power this entire system.