Build Your AI-Powered Business Brain: Obsidian + Claude Code

Stop uploading files to ChatGPT. Start building a knowledge system that actually scales.

Vibe coding was the training wheels. Vibes build demos. Sovereigns build businesses. This tutorial is where the shift begins.

Why You Should Switch from Custom GPTs / Claude Projects

If you're using Custom GPTs or Claude Projects to give AI context about your business, you've hit the ceiling. Here's what you're running into — and what you're leaving on the table.

This is Principle 6 — Platform Independence in action: own every critical layer. No vendor can hold your business hostage. See the Sovereign Coding methodology.

The Real Difference

Custom GPTs and Claude Projects are filing cabinets with a chatbot on top. You stuff files in, ask questions, get answers. That's it.

Claude Code + Obsidian is a business operating system. Your AI doesn't just know things — it can do things. It searches your knowledge base, follows your processes, writes in your voice, creates documents, and executes workflows. And every piece of knowledge you add makes it smarter, permanently.

I run 268 specialized AI agents out of my Obsidian vault. Each one knows exactly what it's responsible for, what knowledge to reference, and how to interact with other agents. Try building 268 Custom GPTs and keeping them all in sync.

Part 1: Getting Started (15 Minutes)

Install the Tools

You'll need:

• A Claude subscription (Pro at $20/mo or Max at $100/mo)
• Node.js 18+ installed (for Claude Code)

Create Your Vault

1. Open Obsidian → "Create new vault"
2. Name it something like Business-Brain or [YourCompany]-KB
3. Pick a location: ~/Documents/Business-Brain/

Set Up the Folder Structure

Business-Brain/
├── CLAUDE.md                  ← YOUR AI'S BRIEFING DOCUMENT
├── Knowledge-Base/
│   ├── company/
│   │   ├── about.md
│   │   ├── services.md
│   │   ├── ideal-client.md
│   │   └── brand-voice.md
│   ├── processes/
│   │   ├── sales-process.md
│   │   ├── onboarding.md
│   │   └── fulfillment.md
│   ├── marketing/
│   │   ├── content-strategy.md
│   │   ├── email-templates.md
│   │   └── social-media-playbook.md
│   └── industry/
│       ├── competitor-analysis.md
│       ├── market-trends.md
│       └── compliance-notes.md
├── Agents/
│   ├── TEMPLATE.md
│   ├── content-strategist.md
│   ├── sales-outreach.md
│   └── operations-writer.md
├── Clients/
│   └── [client-name].md
├── Meetings/
│   └── [date]-[topic].md
└── Templates/
    ├── client-profile.md
    ├── meeting-notes.md
    └── proposal.md

Write Your CLAUDE.md

This is the single most important file. Claude Code reads it automatically every time you start a session in this folder. Think of it as your AI's employee handbook.

Create CLAUDE.md in the root of your vault:

# CLAUDE.md — [Your Company] Business Brain

## What This Is
This vault is the knowledge base and agent system for [Your Company].
It contains everything you need to help me run my business: our services,
processes, brand voice, client info, and specialized AI agents.

## About the Business
- **Company**: [Your Company Name]
- **Industry**: [Your Industry]
- **Owner**: [Your Name]
- **What we do**: [One-liner description]
- **Website**: [URL]

## How to Help Me

### When I ask you to write content:
1. Read `Knowledge-Base/company/brand-voice.md` for tone and style
2. Reference `Knowledge-Base/company/services.md` for accurate descriptions
3. Check `Knowledge-Base/company/ideal-client.md` to tailor the message

### When I ask about a client:
1. Search `Clients/` for their file
2. Check `Meetings/` for recent notes

### When I ask about a process:
1. Check `Knowledge-Base/processes/` for the relevant SOP

### When I reference an agent:
1. Find the agent file in `Agents/`
2. Read it and adopt that agent's identity, knowledge, and approach
3. Reference the knowledge base paths listed in the agent's metadata

## Writing Rules
- [Your tone: e.g., "Professional but conversational. No corporate jargon."]
- [Your terminology: e.g., "We say 'clients' not 'customers'"]
- [Your CTA style: e.g., "Every piece of content ends with a clear next step"]
- [Your no-go list: e.g., "Never use the words 'synergy' or 'leverage'"]

## Current Priorities
- [e.g., Launching new group coaching program by June 1]
- [e.g., Building email list to 5,000 subscribers]
- [e.g., Closing 3 enterprise deals this quarter]

Your CLAUDE.md is the blueprint for how the AI works in your vault. This is Principle 1 — Blueprint Before Build: nothing gets built until it's been drawn, spec'd, or PRD'd. See the Sovereign Coding methodology.

Test It

Open your terminal:

cd ~/Documents/Business-Brain
claude

Try a prompt:

"Draft a follow-up email to a prospect who attended our workshop last week. Match our brand voice."

Claude reads your CLAUDE.md, finds your brand voice file, checks your services, and writes something that actually sounds like you.

Part 2: Building Your Knowledge Base

Your Knowledge Base is where you store everything Claude needs to know about your business. The more you put in, the smarter it gets.

What Goes in the Knowledge Base

Writing Effective Knowledge Base Entries

The best KB entries are structured and specific. Claude can search and reference them, so clear headings matter.

Good example — Knowledge-Base/company/services.md:

# Our Services

## Executive Coaching (Flagship)
- **Price**: $5,000/quarter
- **Duration**: 12-week engagement, weekly 1-hour calls
- **Ideal for**: Business owners doing $1M-$10M who feel stuck
- **Key outcome**: Clear 90-day plan + accountability system
- **Differentiator**: We combine business strategy with personal performance

## Group Mastermind
- **Price**: $1,500/quarter
- **Duration**: Ongoing, monthly in-person + weekly virtual
- **Ideal for**: Entrepreneurs who want peer accountability
- **Max size**: 12 members
- **Differentiator**: Application-only, curated by industry

## Strategy Intensive (Entry Point)
- **Price**: $500 one-time
- **Duration**: Half-day deep dive
- **Ideal for**: Prospects considering coaching
- **Key outcome**: Custom growth roadmap document
- **Upsell path**: Executive Coaching

Why this works: When you tell Claude "draft a proposal for a prospect interested in coaching," it pulls the exact pricing, outcomes, and differentiators — not generic fluff.

Use Frontmatter for Structure

Add YAML frontmatter to make entries machine-searchable:

---
type: process
category: sales
status: active
last_updated: 2026-04-15
---

# Sales Process

## Stage 1: Discovery Call
...

Structured frontmatter isn't decoration — it's the spec the AI reads. This is Principle 2 — Specs Over Vibes: your PRD is your contract with the AI. Specificity in equals quality out. See the Sovereign Coding methodology.

Build It Incrementally

You don't need to write everything day one. Add knowledge as you work:

• After a great client call: Add notes to Meetings/ and update the client file
• After you nail a process: Document it in processes/
• After you write killer copy: Save it in marketing/
• After you learn something about your market: Add it to industry/

Every entry you add is permanent leverage. Claude remembers it across every future session.

This is Principle 7 — Compounding Knowledge: every session ends with capture. Nothing learned is ever re-learned. After a year, you're not building with an AI; you're building with your AI, loaded with every pattern you've ever shipped. See the Sovereign Coding methodology.

Part 3: Building Your Agent Wiki

This is where it gets powerful. Agents are specialized AI personas stored as markdown files. Instead of one generic AI assistant, you build a team of specialists.

What Is an Agent?

An agent file is a markdown document that tells Claude:

• Who this agent is (identity and expertise)
• What it knows (which KB entries to reference)
• How it works (interaction style, process, output format)

When you say "use the content strategist agent," Claude reads that file and becomes that specialist for the conversation.

Agent Template

Save this as Agents/TEMPLATE.md:

---
name: [Agent Name]
type: [marketing|sales|operations|strategy|creative]
status: active
last_updated: [date]

trigger_phrases:
  - "keyword one"
  - "keyword two"
  - "specific use case"

knowledge_base:
  primary: Knowledge-Base/[relevant-folder]/
  additional:
    - Knowledge-Base/company/brand-voice.md
    - Knowledge-Base/company/services.md
---

> **Frontmatter fields:** `trigger_phrases` defines what activates the agent; `knowledge_base` defines what context Claude pulls in.

# [Agent Name]

## Identity
[2-3 sentences: Who is this agent? What's their expertise?
What makes them different from a generic AI?]

## Core Capabilities
- **Capability 1**: What it does and when to use it
- **Capability 2**: Another key function
- **Capability 3**: Specialized skill

## Process
1. **Step 1**: How the agent starts (gather info, assess situation)
2. **Step 2**: How it works (analysis, creation, strategy)
3. **Step 3**: What it delivers (output format, next steps)

## Rules
- [Specific guidelines this agent follows]
- [Tone, format, or compliance requirements]
- [What this agent does NOT do — clear boundaries]

## Output Format
[Describe what the deliverable looks like.
Bullet points? Long-form? Structured template? Email draft?]

Three Starter Agents (Copy These)

This is Principle 4 — Orchestrate, Don't Improvise: specialized agents for specialized work. Don't ask one generalist AI to do everything. Vibe coders have one hammer; Sovereigns have a full workshop with a foreman. See the Sovereign Coding methodology.

1. Content Strategist

Save as Agents/content-strategist.md:

---
name: Content Strategist
type: marketing
status: active
last_updated: 2026-04-15

trigger_phrases:
  - "content strategy"
  - "content plan"
  - "what should I post"
  - "content calendar"
  - "blog ideas"
  - "social media plan"

knowledge_base:
  primary: Knowledge-Base/marketing/
  additional:
    - Knowledge-Base/company/brand-voice.md
    - Knowledge-Base/company/ideal-client.md
    - Knowledge-Base/company/services.md
---

# Content Strategist

## Identity
Senior content strategist specializing in thought leadership content
for service-based businesses. Thinks in terms of content pillars,
audience awareness stages, and conversion paths — not just "what
to post today."

## Core Capabilities
- **Content Calendar Creation**: Weekly/monthly plans tied to business goals
- **Topic Ideation**: Generate content ideas from existing knowledge, industry trends, and client pain points
- **Repurposing Strategy**: Turn one piece of content into 5-10 assets across platforms
- **Performance Analysis**: Evaluate what's working and recommend adjustments

## Process
1. **Assess**: Review current business priorities, upcoming launches, and target audience
2. **Plan**: Create content themes tied to business objectives
3. **Create**: Draft content briefs or full pieces using brand voice guidelines
4. **Distribute**: Map content to platforms with format adaptations

## Rules
- Always tie content back to a business objective (lead gen, authority, nurture)
- Match the brand voice in Knowledge-Base/company/brand-voice.md
- Every content piece should have a clear call to action
- Prioritize depth over frequency — one great piece beats five mediocre ones

## Output Format
Content plans as structured tables with: Date, Platform, Topic, Hook, CTA, Content Pillar

2. Sales Outreach Specialist

Save as Agents/sales-outreach.md:

---
name: Sales Outreach Specialist
type: sales
status: active
last_updated: 2026-04-15

trigger_phrases:
  - "cold outreach"
  - "follow up email"
  - "sales email"
  - "prospect outreach"
  - "lead nurture"
  - "drip sequence"

knowledge_base:
  primary: Knowledge-Base/processes/sales-process.md
  additional:
    - Knowledge-Base/company/services.md
    - Knowledge-Base/company/ideal-client.md
    - Knowledge-Base/marketing/email-templates.md
---

# Sales Outreach Specialist

## Identity
Direct-response sales copywriter who crafts personalized outreach
that feels human, not templated. Specializes in warm and cold
outreach for service-based businesses where trust is the currency.

## Core Capabilities
- **Cold Outreach Sequences**: 3-5 email sequences for new prospects
- **Follow-Up Cadences**: Timely, relevant follow-ups that don't feel desperate
- **Personalization at Scale**: Frameworks for personalizing outreach using available data
- **Objection Handling**: Response templates for common pushback

## Process
1. **Research**: Review prospect info, find connection points
2. **Craft**: Write outreach using the problem→solution→proof→CTA framework
3. **Sequence**: Build multi-touch sequences with varied angles
4. **Refine**: A/B test subject lines and opening hooks

## Rules
- Never sound like a template. Every email should feel written for one person.
- Lead with their problem, not your product
- Keep emails under 150 words for cold outreach
- Include exactly one CTA per email
- Reference the sales process in Knowledge-Base/processes/sales-process.md

## Output Format
Emails with: Subject Line, Body, CTA, Send Timing, Sequence Position

3. Operations/SOP Writer

Save as Agents/operations-writer.md:

---
name: Operations & SOP Writer
type: operations
status: active
last_updated: 2026-04-15

trigger_phrases:
  - "write an SOP"
  - "document this process"
  - "create a procedure"
  - "standard operating procedure"
  - "how do we do"
  - "process documentation"

knowledge_base:
  primary: Knowledge-Base/processes/
  additional:
    - Knowledge-Base/company/about.md
---

# Operations & SOP Writer

## Identity
Operations specialist who turns messy, ad-hoc processes into
clear, repeatable SOPs that anyone on the team can follow.
Thinks in systems — every process should be delegatable.

## Core Capabilities
- **Process Documentation**: Turn verbal explanations into structured SOPs
- **Workflow Design**: Map out multi-step processes with decision trees
- **Checklist Creation**: Convert complex processes into actionable checklists
- **Gap Analysis**: Identify missing steps, edge cases, and failure points in existing processes

## Process
1. **Extract**: Ask questions to understand the full process
2. **Structure**: Organize into numbered steps with clear ownership
3. **Edge Cases**: Identify "what if" scenarios and document how to handle them
4. **Format**: Deliver in a consistent, scannable format

## Rules
- Every step must have a clear action verb ("Send the email" not "Email should be sent")
- Include estimated time per step where possible
- Note which steps require human judgment vs. can be automated
- Link to related processes in Knowledge-Base/processes/

## Output Format
Structured SOP with: Title, Purpose, Scope, Prerequisites, Steps (numbered),
Edge Cases, Tools Required, Estimated Time

How to Use Your Agents

In a Claude Code session:

"Read the content strategist agent and create a content plan for next month.
Our priority is launching the group coaching program."

Claude reads the agent file, adopts the content strategist persona, references your KB for brand voice and services, and produces a structured content calendar.

"Use the sales outreach agent to write a 3-email cold sequence for
agency owners who might need our consulting services."

"Use the SOP writer agent to document our client onboarding process.
Here's how it works: [describe it verbally]"

How Agents Reference Your Knowledge Base

This is the key architectural pattern. Notice that each agent has a knowledge_base section in its frontmatter:

knowledge_base:
  primary: Knowledge-Base/marketing/
  additional:
    - Knowledge-Base/company/brand-voice.md
    - Knowledge-Base/company/services.md

This tells Claude exactly which parts of your vault to read when using that agent. Your agents get smarter as your Knowledge Base grows — without you ever updating the agent file.

Connecting It All in CLAUDE.md

Add this to your CLAUDE.md:

## Agent System

When I reference an agent by name or describe a task that matches
an agent's trigger phrases:
1. Find the matching agent file in `Agents/`
2. Read it fully — adopt the identity, capabilities, and rules
3. Read the knowledge base paths listed in the agent's metadata
4. Execute the task using that agent's process and output format

Available agents:
- **Content Strategist** → content planning, ideation, calendars
- **Sales Outreach** → cold emails, follow-ups, sequences
- **Operations Writer** → SOPs, process documentation, checklists

Part 4: Power Moves

1. Client Intelligence Files

Every client gets a structured file in Clients/:

---
type: client
status: active
value: high
last_contact: 2026-04-15
---

# [Client Name] — [Company]

## Background
- How we met: [Referral, conference, cold outreach]
- Their business: [What they do, size, stage]
- Key pain points: [Why they came to us]

## Engagement History
- [Date]: Started with Strategy Intensive
- [Date]: Upgraded to Executive Coaching
- [Date]: Referred two contacts

## Communication Notes
- Prefers email over calls
- Decision maker, doesn't need to check with partners
- Responds quickly to direct, concise messages

## Opportunities
- Potential fit for Group Mastermind
- Could be a case study if results continue

Now when you say "draft a check-in email to [Client]," Claude knows their history, preferences, and opportunities.

2. Meeting Notes That Compound

# 2026-04-15 — Call with [Name]

## Context
[Why this meeting happened]

## Key Points
- [What was discussed]
- [Decisions made]

## Action Items
- [ ] [What you need to do]
- [ ] [What they need to do]

## Follow-up
- Date: [When]
- Method: [Email/Call/Meeting]
- Topic: [What to follow up on]

3. Templates Folder

Store your best templates so Claude can use them as starting points:

Templates/
├── proposal-template.md
├── weekly-newsletter.md
├── case-study-format.md
├── onboarding-checklist.md
└── meeting-notes.md

4. Claude Code's Built-in Memory

Claude Code reads a CLAUDE.md file in your project root on every session — anything in it persists across conversations as standing instructions.

To add a fact mid-conversation, prefix your message with # (or use the /memory command):

• # We never discount below 20%
• # Sarah handles all scheduling
• # Our Q3 focus is enterprise clients

Claude Code appends each entry to the appropriate CLAUDE.md (project, user, or local) and applies it in every future session. Plain "remember X" statements in a regular prompt won't persist — you have to use the # prefix or the /memory command for it to write to disk.

5. Growing Your Agent Team

Once you see the pattern, you'll want more agents. Common ones:

Each one follows the same template. Create the file, point it at the right knowledge, define its rules.

Quick-Start Checklist

• Install Obsidian (free) and Claude Code ($20-100/mo)
• Create your vault with the folder structure above
• Write your CLAUDE.md — this is the foundation
• Add core knowledge: about.md, services.md, brand-voice.md
• Copy the 3 starter agents into your Agents/ folder
• Open terminal, cd to your vault, run claude
• Test: "Read the content strategist agent and give me 10 content ideas for this month"
• Add one client file, one process doc, one meeting note
• Watch it get smarter every time you add knowledge

FAQ

"I'm not technical. Can I do this?" If you can create folders and write in a text editor, you can do this. Obsidian is designed for non-technical people. The only terminal command you need is cd [folder] && claude.

"How is this different from just chatting with Claude?" When you chat with Claude normally, it knows nothing about your business. Every conversation starts from zero. With this system, Claude reads your CLAUDE.md automatically and can search your entire knowledge base. It's the difference between hiring a temp every day vs. having a full-time employee who knows your business.

"Do I need to know how to code?" No. Everything here is plain markdown — basically text with some formatting. No programming required.

"Is my data safe?" Your files stay on your local machine. When you interact with Claude Code, the content of files you reference is sent to Anthropic's API for processing — same as pasting text into any Claude conversation. It is not used for training. Review Anthropic's privacy policy for details.

"What if my vault gets huge?" Claude Code doesn't load your entire vault into memory. It searches for relevant files when needed. Thousands of notes work fine.

"Can I share agents with my team?" Yes. Agent files are just markdown. You can email them, put them in a shared folder, or use git to sync them. This is one of the biggest advantages over Custom GPTs — your agents are portable.

"Can I still use Obsidian normally?" Absolutely. Obsidian is your reading/writing interface. Claude Code works with the same files from the terminal. They're complementary — Obsidian for browsing and organizing, Claude Code for searching and building.

"Can Claude edit my Obsidian notes?" Yes. Claude Code can create new files and edit existing ones. It can update client records, create meeting notes, append to your KB — all within your vault.

What's Next: Sovereign Coding™

If you finished this tutorial and built a working Business Brain, you just practiced five of the seven principles of a methodology I call Sovereign Coding™ — the structured, disciplined evolution of vibe coding for solopreneurs who ship software with AI but don't write code themselves.

The 7 Principles:

1. Blueprint Before Build — Diagram/PRD before code
2. Specs Over Vibes — The PRD is the contract with AI
3. Proof Over Promises — No stubs, actual testing, embarrassment test
4. Orchestrate, Don't Improvise — Specialist agents for specialist work
5. Pattern → Product — Every project becomes reusable IP
6. Platform Independence — Own every critical layer
7. Compounding Knowledge — Every session ends with capture

This tutorial is Principles 1, 2, 4, 6, and 7 applied to knowledge work. Sovereign Coding teaches Principles 3 (Proof Over Promises) and 5 (Pattern → Product) applied to shipping production software — turning this Business Brain into a Business.

You're not a developer who can't read code. You're an architect who doesn't need to.

Questions? Hit me up in the Driven group.

Resources

• Claude Code docs: https://docs.anthropic.com/en/docs/claude-code
• Obsidian docs: https://help.obsidian.md
• Sovereign Coding methodology: 7 principles in full

Built by Matt Nye | April 2026