Diagnostic: Why are you stuck? Take the 7-minute audit. Start Quiz →

Frank Anaya — The Practical Tutorial · March 2026

The Claude Tutorial

Cowork, Obsidian, Skills & the Full Ecosystem

Not another overview. This is the step-by-step tutorial — what to type, what happens, and why it matters. From your first session to a custom AI system that knows your voice, your vault, and your workflow.

All Skill Levels

Claude Code Claude Cowork Obsidian Skills CLAUDE.md MCP

In This Tutorial

The Claude Ecosystem — What Goes Where

There is not one “Claude.” There are four interfaces to the same underlying intelligence, each designed for a different moment in your work. Understanding which to use when is the first decision that separates people who get value from AI from people who are just chatting with it.

The Chatbot

Claude.ai

Browser-based conversation. Paste content in, get responses. No file access. No memory between sessions. Great for quick questions, brainstorming, and one-off drafts where you’ll do the implementation yourself. Best for: Quick Q&A, brainstorming, ad hoc drafts

The Collaborator

Claude Code

Terminal-based AI that reads your actual files, edits them directly, remembers your preferences via CLAUDE.md, and executes multi-step tasks. This is the main tool this tutorial covers. Best for: Real work on real files — writing, PKM, vault management

The Desktop App

Claude Desktop

Native Mac/Windows application with built-in MCP support. Connects to local tools (Obsidian, browsers, databases) through the same protocol Claude Code uses, but with a graphical interface instead of a terminal. Best for: MCP integrations without the terminal

For Developers

Claude API

The raw API for building Claude into your own applications, automations, and custom tools. Requires programming knowledge. Powers everything from custom chatbots to automated content pipelines. Best for: Building custom tools and integrations

The Decision Framework

Which Claude Tool for Which Moment

This is the decision most people never make explicitly. They pick one tool and use it for everything, or they switch randomly. Here is how to decide deliberately.

Quick question, no files involved → Claude.ai

Need AI to read or edit my actual files → Claude Code

Want to delegate a task and keep working → Claude Cowork

Need AI connected to Obsidian, browser, or other apps → Claude Code + MCP or Claude Desktop

Want a repeatable workflow I can trigger with one command → Claude Code + Custom Skill

Building my own app or automation pipeline → Claude API

The power move is using multiple tools in the same day. Claude.ai for morning brainstorming. Claude Code for editing your draft. Cowork for a vault audit running in the background while you write. Skills to trigger your weekly review with one command. That is a system — not just a tool.

"The system you trust is the one you actually use. Build it around how you already work — not how a tool wants you to work."

Prerequisite

Setup — Terminal to First Session in 5 Minutes

If you already have Claude Code installed, skip to Section 4: Claude Cowork. If not, here is the complete setup — no assumed knowledge.

Install Node.js

Go to nodejs.org and download the LTS version for your operating system. Install it. You do not need to understand what Node.js is — it is the runtime that lets Claude Code run on your machine. Takes about two minutes.

Install Claude Code

Open your terminal (Mac: search “Terminal” in Spotlight. Windows: use WSL) and run one command:

$ npm install -g @anthropic-ai/claude-code
Navigate to the folder you want to work in
$ cd ~/Documents/MyObsidianVault
Launch Claude Code
$ claude
First thing: create your memory file
> /init

Authenticate

After running claude, your browser opens to Anthropic’s login page. Sign in with the account that has your Pro or Max subscription. Done — you are now in a Claude Code session.

Create your CLAUDE.md

Type /init in the session. Claude creates a CLAUDE.md file with starter content. This is your persistent memory — edit it now or come back to it. Every future session in this folder starts by reading it.

The Game Changer

Claude Cowork — Delegate and Keep Working

Most people use Claude Code interactively: you type a prompt, Claude works, you wait, you review, you type the next prompt. Cowork changes this entirely. You assign Claude a task, and it runs in the background while you continue your own work. When it is finished, it notifies you.

This is the difference between having an assistant who can only work when you are watching and one who can go off, do the work, and come back with the result. For knowledge workers, this is where Claude stops being a tool you use and starts being a teammate who handles things.

How Cowork Actually Works

Start a Cowork session

From any Claude Code session, launch Cowork with a task description. Claude begins working autonomously — reading files, making edits, creating new documents — while you switch to something else.

Claude works in the background

Cowork uses the same capabilities as Claude Code — file reading, editing, creating, searching — but runs them without waiting for your input at each step. It plans, executes, and self-corrects.

You get notified when it’s done

Claude tells you what it did, what it changed, and what it created. You review the output. If something needs adjustment, you give feedback in the same session.

# Navigate to your vault
$ cd ~/Documents/MyObsidianVault
$ claude
Start a Cowork task
> Review all my notes tagged #project-ideas from the past month.
For each one, write a one-paragraph assessment: is this idea
ready to develop, does it need more research, or should I
archive it? Save the assessment as “Project Review - March.md”
in my Reviews folder.
Claude begins working… you can switch to Obsidian, write, do other tasks.
✓ Done. Created “Project Review - March.md” with assessments for 14 project ideas.

What Cowork Is Best For

Cowork shines on tasks that are well-defined, multi-step, and do not require your creative judgment at every step. Think of it as everything you would delegate to a research assistant or virtual assistant — if that assistant could read your entire vault.

Use Interactive Claude Code When

• You need creative input at each step
• The task requires back-and-forth judgment
• You want to steer the direction in real time
• Drafting content where voice matters deeply
• Exploring ideas — not executing a plan

Use Cowork When

• The task is well-defined with a clear outcome
• It involves scanning many files or notes
• You would normally delegate it to an assistant
• It is time-consuming but not creatively complex
• You want to keep writing while Claude works

Free Download

Get the Creator Vault — Skills, Templates & CLAUDE.md Starters

Includes ready-made skills, CLAUDE.md templates, prompt library, and the Obsidian CLI Quick Reference — everything from this tutorial, organized and ready to use.

Free. No spam. Unsubscribe any time.

The Deep Dive

Claude + Obsidian — The Complete Integration

Your Obsidian vault is already a folder of Markdown files. Claude Code can read and edit all of them the moment you point it at the folder. No plugin required. No setup required. That is level one. But there are three levels — and each one unlocks dramatically more value.

Level 1: Direct File Access (Immediate — No Setup)

Navigate to your vault folder and launch Claude Code. It can now read every note, edit content, create new notes, reorganize folders, and search across your entire vault in plain English.

$ cd ~/Documents/MyObsidianVault
$ claude
Ask about your vault
> What are the main topics across my notes? Which areas have
the most writing, and where are the gaps?
Edit a specific note
> Read “Inbox/Raw idea about attention.md” and turn it into
a proper evergreen note with a clear title, clean structure,
and links to related notes you find in my vault.
Batch operations
> Find all notes in my Inbox folder that have been there
for more than 2 weeks. For each one, either move it to the
right folder based on its content, or mark it for archival
if it seems stale. Show me what you plan to do before doing it.

Level 2: MCP Integration (10 Minutes Setup)

MCP (Model Context Protocol) connects Claude Code to Obsidian’s running instance — not just the files, but the application itself. This gives Claude access to Obsidian’s search engine, tag index, backlink graph, and plugin ecosystem.

# Step 1: In Obsidian
# Settings → Community Plugins → Browse → search "MCP"
# Install the Obsidian MCP plugin → Enable it
Step 2: In Claude Code
$ claude
Add the Obsidian MCP server
> /mcp
Select: Add server → Name: obsidian → URL: http://localhost:22360/sse
Step 3: Test the connection
> How many notes are in my vault? What are my most-used tags?
Your vault contains 1,247 notes. Top tags: #pkm (89), #writing (67), #project (45)…
Now Claude can do things the file system alone can’t:
> Find all notes tagged #idea that have no backlinks pointing to them.
> Show me the graph neighborhood around “Second Brain Framework.md”

Level 1 — File Access Only

• Read and edit note content
• Search by file content
• Create and move files
• Update frontmatter
• Works immediately, no setup

Level 2 — MCP Connection

• Everything in Level 1, plus:
• Search by Obsidian tags natively
• Query backlinks and graph relationships
• Access Obsidian’s search index
• Interact with installed plugins

Level 3: Skills + CLAUDE.md (The Full System)

Levels 1 and 2 give Claude access to your vault. Level 3 gives it understanding. When you combine vault access with a well-written CLAUDE.md (your memory file) and custom skills (your reusable workflows), Claude does not just read your notes — it knows your system, your conventions, your voice, and your goals.

# Your vault's CLAUDE.md — loaded every session automatically
My Vault Structure

Inbox/ — raw captures, unprocessed
Notes/ — evergreen notes, developed thinking
Projects/ — active project notes
Reviews/ — weekly and monthly reviews
Templates/ — my note templates

My Conventions

New notes start as a file in Inbox/
Developed notes move to Notes/ with [[backlinks]] added
I use YAML frontmatter: tags, created, status
Status values: seedling, growing, evergreen
I write in first person, present tense

When Working With My Notes

Never delete a note — move to Archive/ if stale
Always add at least two [[backlinks]] to related notes
Match my writing voice (direct, warm, no jargon)
Ask before reorganizing more than 5 files at once

With this CLAUDE.md in place, every prompt you type or skill you run already knows your folder structure, your naming conventions, your status system, and your rules. Claude does not just have access to your vault — it has context about how your vault works. That is the difference between a tool that reads files and a system that understands your workflow.

The Memory Layer

CLAUDE.md Masterclass — Beyond the Basics

If you have read the Claude Code guide, you know CLAUDE.md exists. This section goes deeper — what separates a mediocre CLAUDE.md from one that genuinely transforms every session.

The Anatomy of an Excellent CLAUDE.md

Most people write a vague CLAUDE.md: “I’m a writer. I like clarity. Don’t use jargon.” This is the equivalent of handing a new employee a one-sentence job description. Here is what actually works — organized by what matters most.

Section 1: Identity & Context

Tell Claude who you are, what you do, and what this specific project is about. Be specific enough that Claude could explain your work to someone else accurately.

MD CLAUDE.md — Identity section

# Who I Am

Frank Anaya. I write about personal knowledge management, creative systems, and using AI as a thinking partner — not a replacement for thinking. My audience is professionals who capture everything but struggle to ship finished work. I call them “9-to-5 Architects.”

# This Project

This is my Obsidian vault. It contains ~1,200 notes organized into Inbox (raw captures), Notes (developed thinking), Projects (active work), and Reviews (weekly/monthly reflections). My current focus is developing a course about building creative systems.

Section 2: Voice & Style

This is where most CLAUDE.md files fail. “Write in my voice” means nothing to Claude. Give it measurable, observable patterns from your actual writing.

MD CLAUDE.md — Voice section

# My Writing Voice

• Short sentences. Rarely more than 20 words. Fragments are fine.
• First person, present tense by default
• Direct and warm — like a mentor, not a marketer
• I use dashes — like this — more than parentheses
• I avoid: exclamation points, “amazing”, “incredible”, “game-changer”, passive voice
• I use occasional dry humor, never forced
• When referencing tools or methods, I explain what they do, not what they’re called

Test: If you write something and it could appear on any generic blog, rewrite it. My writing should sound like only I would say it that way.

Section 3: Rules & Boundaries

These are your non-negotiables — the things that, if Claude violates them, the output is wrong regardless of how well-written it is.

MD CLAUDE.md — Rules section

# Always

• Read existing files before editing — understand the context first
• Ask before making changes to more than 5 files
• When creating notes, add YAML frontmatter with tags, created date, and status
• Add at least two [[backlinks]] to related notes when developing any note
• Summarize what you did after completing any multi-step task

# Never

• Never delete any file — move to Archive/ instead
• Never add ideas I did not express in my notes — refine and develop, do not invent
• Never use generic AI filler phrases (“In today’s fast-paced world…”)
• Never restructure my folder hierarchy without showing me the plan first

The Three-Level Memory Stack

CLAUDE.md is actually three files in a hierarchy. Each level adds context that the levels above it cannot see:

# Level 1: Global — applies to every project you open
~/.claude/CLAUDE.md
  Your identity, universal voice rules, things Claude should always/never do
  Example: "I am Frank Anaya. My writing is direct and warm. Never use passive voice."
Level 2: Project — applies only to this folder
~/Documents/MyVault/CLAUDE.md
Vault structure, current goals, project-specific conventions
Example: “This vault uses STATUS: seedling/growing/evergreen in frontmatter.”
Level 3: Auto-memory — Claude’s own notes
~/.claude/projects/[project]/memory/
Claude remembers things across sessions: decisions made, preferences discovered
Example: “User prefers bullet lists over numbered lists for review notes.”

The strategy: put universal preferences in the global CLAUDE.md. Put project-specific structure and goals in each project’s CLAUDE.md. Let auto-memory handle the small things Claude discovers about you over time. You maintain the first two. Claude maintains the third.

The Power Feature

Building Custom Skills from Scratch

A skill is a Markdown file that becomes a slash command. You write the instructions once in plain English, save the file, and it appears in your Claude Code session as a command you can trigger instantly. No code. No configuration. No compilation step.

If CLAUDE.md is your memory, skills are your habits — the repeatable workflows you want to run the same way every time without re-explaining them.

Skill File Anatomy

A skill is a Markdown file saved at .claude/skills/skill-name.md inside your project folder. The filename (minus the .md extension) becomes the slash command. Here is the structure:

Skill .claude/skills/morning-review.md

---

name: Morning Review

description: Read my Daily Notes from the past week and generate a morning briefing with priorities, open threads, and recurring themes.

---

Instructions

Read all Daily Notes from the past 7 days (they are in the DailyNotes/ folder, named YYYY-MM-DD.md).

Generate a morning briefing with these sections:

• Focus areas this week — what topics and projects keep appearing
• Unfinished threads — tasks or ideas mentioned but not resolved
• Recurring themes — patterns in my thinking I might not notice day-to-day
• Today’s top 3 priorities — based on what is most urgent and unfinished

Save the briefing as “Morning Briefing - [today’s date].md” in the Reviews/ folder.

Keep the tone direct and concise. Do not add motivational filler. If there is nothing notable in a section, say so — do not pad it.

# In any Claude Code session in your vault:
> /morning-review
Reading 7 Daily Notes… analyzing themes… generating briefing…
✓ Created “Morning Briefing - 2026-03-19.md” in Reviews/

The Three Parts Every Skill Needs

Frontmatter — The What and Why

The name is what appears in the slash command menu. The description tells Claude (and you) what this skill does. Keep the description specific — it helps Claude understand the intent when you invoke it.

Instructions — The How

Write the step-by-step instructions in plain English. Be specific about: what files to read, what to produce, where to save it, and what format to use. The more concrete you are, the more consistent the output.

Constraints — The Guardrails

Tell Claude what not to do. “Do not add ideas I did not express.” “Keep it under 500 words.” “Do not reorganize my folder structure.” Constraints prevent the skill from drifting off-target over time.

Skills That Accept Arguments

Skills can take input when you invoke them. Reference the argument in your instructions and Claude fills it in at runtime.

Skill .claude/skills/develop-note.md

---

name: Develop Note

description: Take a rough note from the Inbox and develop it into a structured evergreen note.

---

The user will provide a note title or filename as an argument.

Read the specified note. Then:

• Rewrite the title to be clear and specific (not a question, not a date)
• Structure the content with a clear opening claim, supporting points, and connections
• Add YAML frontmatter: tags (infer from content), created (today), status: growing
• Find 2-3 related notes in the vault and add [[backlinks]]
• Move the file from Inbox/ to Notes/
• Keep my voice. Develop my thinking — do not add ideas I did not have

> /develop-note "Inbox/raw thoughts on digital minimalism.md"
Reading note… developing structure… finding related notes…
✓ Developed and moved to Notes/Digital Minimalism as Creative Constraint.md
Added links to: [[Attention as Currency]], [[The Collector’s Trap]]

Copy, Paste, Use

6 Skills You Can Copy Right Now

Each skill below is production-ready. Create the file at the path shown, paste the content, and the skill appears in your slash command menu immediately. Modify the instructions to match your vault structure and conventions.

Daily Practice

/morning-review

Reads your Daily Notes from the past week and generates a morning briefing with focus areas, open threads, and today’s priorities.

File: .claude/skills/morning-review.md

Read all files in DailyNotes/ from the past 7 days. Generate a briefing: what I focused on, what is unresolved, recurring themes, and three priorities for today. Save as “Morning Briefing - [date].md” in Reviews/. Be concise. No filler. If a section has nothing notable, say so.

Content Creation

/draft-newsletter

Turns scattered notes and captures into a coherent newsletter draft in your voice.

File: .claude/skills/draft-newsletter.md

The user provides a topic or folder. Read all relevant notes and captures. Find the most interesting thread across them. Draft an 800-word newsletter: lead with the core insight (not context), develop the argument with specifics from my notes, end with one actionable takeaway. Match my voice from CLAUDE.md. Save as draft in Content/Newsletters/. Do not add ideas I did not express — synthesize and develop, do not invent.

Vault Maintenance

/inbox-process

Processes everything sitting in your Inbox folder — develops, files, or archives each note.

File: .claude/skills/inbox-process.md

Scan every file in Inbox/. For each note, classify it: (A) Has enough substance to develop — move to Notes/ and add frontmatter + backlinks. (B) Belongs in an active project — move to the relevant Projects/ subfolder. (C) Too vague or stale — move to Archive/. Show me your plan for all files before executing. Process one at a time so I can approve each decision.

Weekly Ritual

/weekly-review

Generates a comprehensive weekly review from your Daily Notes and file activity.

File: .claude/skills/weekly-review.md

Read all Daily Notes from the past 7 days and all files modified this week. Generate a weekly review with: (1) Accomplishments — what I finished or made progress on, (2) Open threads — what is unresolved, (3) Theme of the week — the one pattern across my captures, (4) Three questions for next week — based on what I was thinking about. Save as “Weekly Review - [date range].md” in Reviews/. Keep it honest. If I had a scattered week, say so.

Connection Building

/link-audit

Finds orphan notes with no backlinks and suggests connections to strengthen your vault’s network.

File: .claude/skills/link-audit.md

Scan all notes in Notes/ and Projects/. Find every note with zero incoming backlinks (nothing links to it). For each orphan, read its content and suggest 2-3 existing notes it should link to, with a one-sentence reason for each connection. Format as a checklist I can work through. Also flag any clusters of notes on the same topic that should have a Map of Content (MOC) note created for them.

Content Multiplication

/multiply

Takes one piece of content and produces versions for five different platforms.

File: .claude/skills/multiply.md

The user provides a note, draft, or newsletter. Read it and create: (1) A LinkedIn post (150 words, professional but personal, ends with a question), (2) A tweet thread opener (280 chars max, hooks with a surprising claim), (3) A YouTube video outline (5 talking points, 30 seconds each), (4) An email newsletter teaser (3 bullets + CTA), (5) Three pull quotes for social graphics. Save all in Content/Repurposed/[title]/. Match my voice for each platform’s conventions.

Systems Thinking

Chaining It All Together — Real Systems

Individual skills are useful. A system of skills, memory, and delegation patterns is transformative. Here are three complete systems you can build by combining what you have learned.

System 1: The Morning Practice

Every morning, you open your vault and run one command. Claude reads your recent notes, generates a briefing, and surfaces your priorities. Total time: 2 minutes. Value: you start every day with clarity instead of scanning notes and hoping something clicks.

$ cd ~/Documents/MyVault && claude
Your morning ritual — one slash command
> /morning-review
Reading 7 Daily Notes…
✓ Created “Morning Briefing - 2026-03-19.md”
Focus areas: course outline, newsletter backlog, Obsidian plugin research
Top priority: The course outline has been mentioned 4 of 7 days — it wants attention.
Open thread: The “attention as currency” idea from Monday is undeveloped.

System 2: The Content Pipeline

You capture ideas all week. On Friday, you run two skills: one to draft a newsletter from your captures, another to multiply it across platforms. By the end of Friday, you have a newsletter draft and five platform-specific versions — from scattered notes to a full content week in under 30 minutes.

# Step 1: Draft newsletter from this week's captures
> /draft-newsletter Inbox/
✓ Created newsletter draft: “Why Your System Fails at Step Two”
Step 2: Multiply into platform versions
> /multiply “Content/Newsletters/Why Your System Fails at Step Two.md”
✓ Created 5 versions in Content/Repurposed/system-fails-step-two/
→ LinkedIn post (147 words)
→ Tweet thread opener (271 chars)
→ YouTube outline (5 talking points)
→ Email teaser (3 bullets + CTA)
→ 3 pull quotes for social

System 3: The Sunday Reset

Sunday evening. You run two skills and one Cowork task. The weekly review generates automatically. The inbox gets processed. A vault audit runs in the background while you read the review. By the time you finish reading, the audit is done and your vault is tighter. You end the week with a clean system and a clear picture of what happened.

# Step 1: Generate weekly review
> /weekly-review
✓ Created “Weekly Review - Mar 13-19.md”
Step 2: Process inbox while you read the review
> /inbox-process
Found 9 notes in Inbox/. Here’s my plan:
→ Develop: 3 notes (sufficient substance)
→ File: 2 notes (belong in active projects)
→ Archive: 4 notes (stale or too vague)
Proceed? [y/n]
Step 3: Run a link audit in the background
> /link-audit
Scanning 1,247 notes… found 43 orphans. Report saved to Reviews/Link Audit - March.md

"A system is not a tool you use. It is a set of habits that run whether you feel motivated or not. Skills make your AI habits automatic."

For Power Users

Advanced Patterns

Once you have the fundamentals — CLAUDE.md, skills, Cowork, Obsidian integration — these patterns take your system to the next level.

Pattern 1: The Auto-Context Hook

Use a SessionStart hook to inject today’s date, your current priorities, and recent file changes into every Claude Code session automatically. You never have to type “today is March 19” again.

# Set up: run /hooks → SessionStart → add this command
echo “Today is $(date ’+%A, %B %d, %Y’).”
echo “Files modified in the last 24 hours:”
find . -name “*.md” -mtime -1 -type f | head -10

Pattern 2: Skills That Chain Other Skills

A skill can instruct Claude to run other skills as part of its workflow. Your Sunday Reset skill could run /weekly-review, then /inbox-process, then /link-audit — all from one command.

Skill .claude/skills/sunday-reset.md

---

name: Sunday Reset

description: Complete weekly reset — review, inbox processing, and link audit in sequence.

---

Run these three workflows in sequence:

• Weekly Review: Read Daily Notes from the past 7 days and all modified files. Generate the weekly review note with accomplishments, open items, themes, and questions. Save in Reviews/.
• Inbox Processing: Scan every file in Inbox/. Classify each as develop, file, or archive. Show me the plan and wait for approval before executing.
• Link Audit: Scan Notes/ and Projects/ for orphan notes. Generate a connection report with suggested links. Save in Reviews/.

Present the weekly review first so I can read it while you process the inbox. After all three are done, give me a one-paragraph summary of what changed.

Pattern 3: Project-Specific CLAUDE.md Files

If you have multiple projects inside your vault — a course, a newsletter, a book — create a CLAUDE.md in each project subfolder. When you run Claude Code from that subfolder, it loads the project-specific context on top of your vault-level context.

# Your vault-level memory (always loaded)
~/Documents/MyVault/CLAUDE.md
Project-specific context (loaded when you cd into it)
~/Documents/MyVault/Projects/Course/CLAUDE.md
“This is a 6-module course on creative systems. Target: intermediate PKM users.”
“Module structure: Intro → Framework → Tool Setup → Practice → Review → Ship”
~/Documents/MyVault/Projects/Newsletter/CLAUDE.md
“Weekly newsletter. 800 words max. Audience: 9-to-5 Architects.”
“Format: Insight → Story → Principle → One Action. Always end with a question.”

Pattern 4: The Voice Calibration Loop

Run this once a month to keep your AI voice matching sharp:

Go Deeper

Resources & Further Reading

From This Site

Companion Guide Claude Code: The Complete Guide The full overview — what it is, pricing, installation, all slash commands, 8 workflows, and mistakes to avoid. Beginner Guide How to Use Obsidian New to Obsidian? Start here. The foundation makes the Claude integration far more powerful. Advanced Guide The Obsidian CLI Guide Automate your vault from the terminal. Claude Code can call CLI commands directly as part of skills. Pillar Guide Personal Knowledge Management The PKM framework that Claude workflows serve — capture, develop, connect, ship.

Official Resources

Official Docs Claude Code Documentation Setup, slash commands, MCP, hooks, skills, memory, and best practices from Anthropic. Pricing Claude Plans & Pricing Current Pro, Max 5x, and Max 20x pricing — updated directly by Anthropic.

Common Questions

FAQ

What is Claude Cowork and how is it different from Claude Code?

Claude Code is an interactive session — you type, Claude responds, you steer. Cowork is a background mode where you assign a task and Claude works autonomously while you do something else. Same tool, different mode. Use interactive mode when you need creative input at each step. Use Cowork when the task is well-defined and you want to delegate it completely.

Do I need to install anything special to use Claude with Obsidian?

No. Your Obsidian vault is a folder of Markdown files. Claude Code reads them immediately — no plugin required. For deeper integration (tag search, backlinks, graph data), install the Obsidian MCP plugin and connect it to Claude Code. But the basic workflow works out of the box the moment you run claude in your vault folder.

How do I create a skill? Do I need to code?

No coding required. A skill is a Markdown file with instructions written in plain English. Save it at .claude/skills/skill-name.md in your project folder. The filename becomes the slash command. Write what you want Claude to do, where to find files, what to produce, and what to avoid. That is the entire process. The six skills in this tutorial are ready to copy and use.

Can I use multiple skills in sequence?

Yes. You can invoke them one after another in a session, or create a “meta-skill” that runs multiple workflows in sequence (like the Sunday Reset skill in this tutorial). Skills can reference the output of previous steps, so each step builds on the last.

What is the difference between Claude.ai, Claude Code, and Claude Desktop?

Claude.ai is a browser chatbot — paste content in, get responses. Claude Code is a terminal-based collaborator that works on your actual files with persistent memory. Claude Desktop is a native Mac/Windows app with built-in MCP support for connecting to local tools. Same underlying intelligence, different interfaces for different workflows.

How long does it take to set up the full system described here?

Installation: 5 minutes. Basic CLAUDE.md: 15 minutes. Six ready-made skills: 10 minutes to create the files, 20 minutes to customize them for your vault. MCP connection: 10 minutes. Total to a fully working system: about an hour. But you do not need to do it all at once — start with installation and CLAUDE.md, add skills as you discover workflows you repeat.

Is my vault data private? Does Claude store my notes?

Claude Code reads files locally on your machine during the session. Your vault data is sent to Anthropic’s API for processing (like any Claude interaction), but it is not stored or used for training by default. Nothing leaves your computer unless you use MCP to connect external services. For sensitive content, use /permissions to restrict which folders Claude can access.

Which Claude plan do I need for everything in this tutorial?

Claude Pro ($20/month) covers everything in this tutorial — Claude Code, Cowork, skills, MCP, CLAUDE.md. The default model (Sonnet 4.6) handles all these workflows well. Max plans ($100-$200/month) unlock Opus 4.6 and higher usage limits, which matter for very long sessions or processing large vaults. Start with Pro.

— Free

Build Your System. Start Here.

The Creator Vault includes CLAUDE.md starters, ready-made skills, Obsidian templates, and the prompt library — everything pre-built and organized for the workflows in this tutorial.

Free. No spam. Unsubscribe any time.

Published March 19, 2026 Share on X → Get the Newsletter →