Step 1 Build Jarvis inside Obsidian

Step 1 Build Jarvis inside Obsidian Not voice assistant. Personal intelligence system. Reads your notes Runs on schedule Finds patterns Gets smarter every week Foundation first: vault + claude + memory + skills Step 2 This is not a voice assistant It runs in background of your life. Typical assistant Waits for prompt Answers question Easy to forget Jarvis system Reads notes automatically Tracks projects and decisions Surfaces patterns and open loops Improves every week From chat tool to background intelligence layer Step 3 Week 1 -> Week 12 Useful becomes indispensable. 1 Week 1 Morning brief beats manual project recap. 2 Week 4 It connects last month's decision to today's problem. 3 Week 12 It identifies work pattern hiding for 3 months. Compounding context changes output quality. Step 4 Passive vault vs active vault Same notes. Different behavior. Passive You write notes Notes sit there You search later Maybe you find something Active Reads notes automatically Generates outputs on schedule Surfaces missed connections Applies what it learned last week Schedule Every day, 09:00 Scan new notes Update insights Generate output Morning brief 3 new connections 2 follow-ups 1 idea to explore Difference is not storage. Difference is execution. Step 5 Core system architecture Connection + automation + memory. This is what turns notes into an active system. Not plugin stack. Operating loop. Step 6 Why it gets smarter Three compounding mechanisms 1 Accumulated memory Every run writes context to persistent memory. 2 Skill refinement You improve weak outputs and future runs inherit the fix. 3 Pattern recognition Months of notes reveal themes you cannot see manually. Each run leaves residue. Residue improves next run. Step 7 Foundation stack Install core tools first. Claude Desktop Your AI brain and executor Node.js Runtime environment Obsidian Knowledge hub Filesystem MCP File operations & access Hermes Agent Orchestration layer SQLite memory Long-term memory store Telegram notifications Alerts and updates (optional) vault + connection + automation + memory Step 8 Create vault structure One vault. Clear routing. JARVIS-VAULT/ 00-INBOX/ 01-KNOWLEDGE/ 02-PROJECTS/ 03-DAILY/ 04-JARVIS-OUTPUTS/ briefings/ connections/ patterns/ syntheses/ reviews/ 05-RESOURCES/ 06-ARCHIVE/ 07-SYSTEM/ CLAUDE.md memory/ Outputs + system live inside same vault. Step 9 Connect Claude to your vault Use Filesystem MCP. Claude Desktop MCP configuration Config file path --/Library/Application Support/claude_desktop_config.json Configuration { "mcpservers": { "jarvis-vault": { "command": "npx", "args": [ "@modelcontextprotocol/server-filesystem", "-y", "/Users/yourname/JARVIS-VAULT" ] } } } Test: "List all folders in my vault and describe what you see." If Claude sees folder structure, connection is live. Step 10 Add automation layer Install Hermes Agent. > Install Hermes Agent 1 git clone https://github.com/NousResearch/hermes-agent 2 cd hermes-agent 3 npm install Scheduler Runs skills on a schedule. Cron + Natural Retries Handles failures with backoff. 3 attempts HERMES Agent Memory Provides context across runs. Persistent store Outputs Delivers structured results. Summaries + Reports Telegram Sends updates where you are. Bot notifications Hermes runs skills on schedule. Step 11 Configure .env Put memory inside vault. .env MODEL MODEL_PROVIDER=anthropic MODEL_NAME=claude-opus-4-8 ANTHROPIC_API_KEY=your-api-key MEMORY MEMORY_BACKEND=sqlite MEMORY_PATH=/Users/yourname/JARVIS-VAULT/07-SYSTEM/memory/jarvis.db SCHEDULER & OUTPUTS ENABLE_SCHEDULER=true SCHEDULER_TIMEZONE=America/New_York OUTPUT_PATH=/Users/yourname/JARVIS-VAULT/04-JARVIS-OUTPUTS SKILLS SKILL_RETRY_ENABLED=true SKILL_RETRY_MAX=3 SKILL_RETRY_DELAY=300 NOTIFICATIONS NOTIFICATION_GATEWAY=telegram TELEGRAM_BOT_TOKEN=your-bot-token TELEGRAM_CHAT_ID=your-chat-id Memory backs up with vault. Step 12 Write CLAUDE.md This makes Jarvis personal. CLAUDE.md # CLAUDE.md ## 1. Identity ## 2. How I actually work ## 3. Focus areas ## 4. Active projects ## 5. Current beliefs ## 6. Active questions ## 7. Output standards ## 8. Permissions ## 9. Memory rules ## 10. Update protocol Every skill reads this first. 1. Identity Who Jarvis is, who you are. 2. How I actually work Your workflow, rhythm, patterns. 3. Focus areas What matters most right now. 4. Active projects Live projects and context. 5. Current beliefs Principles that guide decisions. 6. Active questions Open questions you're exploring. 7. Output standards Tone, depth, quality bar. 8. Permissions What Jarvis can do (and ask). 9. Memory rules What to remember, what to forget. 10. Update protocol When and how to evolve this file. Best outputs start with best operating doc. Step 13 What goes inside CLAUDE.md Give Jarvis real operating context. Identity Name Role Timezone Work style Best hours Derailers Energizing tasks Focus areas Top priorities Core goals Boundaries Active projects Status Next action Priority Beliefs Principles How you decide What matters Questions Open items Decisions needed Unknowns Decisions Recent calls Rationale Notes Output standards Tone & voice Format Quality bar Permissions What I can do What to avoid Escalate when Write how you actually work. Not idealized version. Step 14 Set boundaries and permissions Autonomy works best with clear rules. Jarvis can do Read any file in vault Write to 04-JARVIS-OUTPUTS Search relevant external info Update memory database Requires approval Write outside outputs folder Contact external parties Financial or strategic recommendations beyond analysis Good Jarvis needs freedom + guardrails. Step 15 Seven skills overview Seven scheduled systems that make Jarvis compound. Morning brief Daily - 6AM Capture processor Daily - 8PM Connection finder Nightly - 11PM Weekly synthesis Sunday - 7PM Belief tracker Monday - 8AM Pattern detector 1st of month Decision intelligence Structure + review decisions Each skill reads context. Each skill writes output. Each skill stores memory. Step 16 Skill 1 Morning brief Daily at 6AM. CLAUDE.md Agent instructions, preferences, context. Yesterday's daily note What happened, outcomes, insights. Active projects Goals, tasks, milestones. Recent briefs Decisions, commitments, open loops. External intelligence News, signals, market, trends. Morning Brief Your daily mission control. THE ONE THING The highest impact priority for today. OPEN LOOPS Commitments, tasks, and decisions pending. PROJECT STATUS Progress, blockers, and next steps. INTELLIGENCE Key signals, updates, and what matters today. DECISION FLAG What requires your attention or choice. Under 5 min. Specific. Vault-grounded. Step 17 Skill 2 Capture processor Daily at 8PM. Today's captures Raw inputs Ideas Tasks Decisions References Insights ... Update CLAUDE.md Keep memory and context fresh. IDEA permanent knowledge TASK project or inbox DECISION decision note REFERENCE resources INSIGHT insight note Prepare tomorrow Review tasks, plan next day. Nothing stays in Captures after processing. Step 18 Skill 3 Connection finder Nightly at 11PM. 1. Recent notes Last 48 hours 2. Vault search Find related history 3. Strong connection test New insight, not same topic Connection note High-signal link discovered. NOTE A <-> NOTE B The two notes being connected. WHAT THIS REVEALS Pattern, relationship, or emergent theme. IMPLICATION Why this matters and what to do with it. SURPRISE LEVEL How unexpected and valuable this link is. Surprising links > obvious overlap. Step 19 Skill 4 Weekly synthesis Sunday at 7PM. Daily notes Raw thoughts, events, and insights. Permanent notes Core ideas, principles, and knowledge. Connection notes Links, relationships, and references. Morning briefs Priorities, signals, and decisions. CLAUDE.md Agent instructions, preferences, context. Weekly synthesis Your week. Clearly understood. Week in one honest line The essence of the week. What advanced Key progress and wins. What stalled Blockers and friction. Belief updates What changed and why. Week's pattern Recurring themes and signals. Next week's single priority One thing that moves it all. Must say what no single note says. Step 20 Skill 5 Belief tracker Monday at 8AM. Current beliefs What I believe right now. Past 7 days of notes Real evidence from my notes. Confirms Supports my current belief. Challenges Contradicts or weakens it. Complicates Adds nuance or new variables. Irrelevant No meaningful connection. Report only what matters Confirmed this week What held up under evidence. Challenged this week What didn't hold up (and why). Severity How significant is the impact on my beliefs. Recommended action What to adjust, test, or stop. Honest challenge beats confirmation bias.