Stop Drowning in Notes! My Brain Is Full Crew Saves Sanity

Stop Drowning in Notes! My Brain Is Full Crew Saves Sanity

Your brain was never designed to hold this much.

Papers, deadlines, half-baked ideas at 2 AM, emails piling up like digital mold, that nagging feeling you've forgotten something critical—again. You've tried Notion. You've tried Roam. You've built elaborate Obsidian vaults that became digital graveyards of good intentions. And still, the overwhelm wins.

What if you never had to organize another note manually again?

What if a team of 8+ AI agents handled your entire second brain—filing, connecting, searching, transcribing, even triaging your email—while you just... talked?

Meet My Brain Is Full Crew, the radical open-source project built by a PhD researcher who hit cognitive bankruptcy and decided to engineer his way out. This isn't another shiny productivity tool for people who already have their lives together. This is a lifeline for the drowning.

In this deep dive, I'll expose why most "second brain" systems fail, how this crew of specialized AI agents actually works under the hood, and exactly how to deploy it in your own Obsidian vault in under 10 minutes. Whether you're a researcher drowning in papers, a developer with brain fog, or anyone who's ever thought "I need to get organized, but I'm too exhausted to get organized"—this might be the system that finally sticks.

What Is My Brain Is Full Crew?

My Brain Is Full Crew is an open-source, multi-platform AI agent system that transforms your Obsidian vault from a static note repository into a living, self-organizing knowledge ecosystem. Built by a PhD researcher experiencing memory decline, dietary chaos, and anxiety-driven overwhelm, it rejects the isolationist approach of traditional second-brain tools.

Here's the critical insight most productivity gurus miss: your brain doesn't operate in a vacuum. Your physical health, mental state, and information diet are inseparable from effective knowledge work. While tools like Notion or bare Obsidian treat organization as purely intellectual, this crew manages three interconnected systems: knowledge capture, nutritional tracking, and mental wellness support.

The project has gained serious traction because it solves a genuinely underserved problem. Most AI+Obsidian integrations are glorified search engines—clever note retrieval for the already-organized. My Brain Is Full Crew is architected for cognitive overwhelm, for people whose mental budgets are depleted and who need externalized executive function, not just externalized memory.

The system runs on four major AI platforms simultaneously: Claude Code, Gemini CLI, OpenCode, and Codex CLI. One codebase deploys everywhere. You clone a single repository, run an installer script, and suddenly your vault has 8 specialized agents and 14 complex skills working in coordinated harmony.

What makes this genuinely different from the flood of AI wrappers flooding GitHub? The chat IS the interface. No dragging files. No manual folder maintenance. No complex tagging rituals. You talk naturally in any language—Italian, French, Japanese, whatever—and the crew executes. The dispatcher routes your intent to the right agent or skill automatically, then chains subsequent operations when work triggers more work.

Key Features That Make This Insanely Powerful

1. Natural Language as Universal Interface

Every interaction happens through conversation. "Save this: meeting with Marco about Q3 budget, report due Friday" becomes a structured note with tasks, wikilinks, and deadline tracking. No syntax to learn, no templates to memorize. The Scribe agent handles the translation from messy human utterance to clean, actionable knowledge.

2. True Multi-Agent Orchestration

This isn't one LLM wearing multiple hats. Eight distinct agents with isolated system prompts, tool restrictions, and model assignments handle specialized functions:

• Architect: Designs vault structure, runs onboarding, establishes governance rules
• Scribe: Transforms stream-of-consciousness dumps into polished notes
• Sorter: Empties your inbox nightly, routes every note to its perfect home
• Seeker: Synthesizes answers across your entire vault with source citations
• Connector: Discovers hidden relationships between notes you'd never spot
• Librarian: Weekly health checks, deduplication, broken link repair, growth analytics
• Transcriber: Converts recordings into structured meeting notes
• Postman: Bridges Gmail/Hey.com and Google Calendar with your vault

3. Dispatcher-Driven Skill System

For complex multi-step workflows, 14 specialized skills run as guided conversations with state maintenance. The /onboarding skill doesn't just create folders—it conducts a full interview about your life, then builds a personalized system. The /email-triage skill scores priority, saves important emails as notes, and chains the Sorter to file them. Skills are checked first by the dispatcher; if no match, quick agent operations handle the rest.

4. Language-Agnostic Operation

Built in English but responds in whatever language you write in. No translation packs, no configuration. The system detects your input language and matches it. This is genuinely rare in AI tooling and reflects the creator's non-native-English perspective.

5. Custom Agent Creation Without Code

Say "create a new agent" and the Architect walks you through designing one from scratch. A budget tracker that monitors spending notes and flags limits. A wardrobe coach that suggests outfits from your notes. A home inventory that prevents your third identical IKEA cutting board. These custom agents coordinate with the core crew, get discovered automatically, and respond in your language.

6. Conservative-By-Default Safety

Agents never delete—only archive. They ask before major decisions. During onboarding, you explicitly accept terms of use. This isn't reckless automation; it's delegation with guardrails.

4 Brutal Real-World Scenarios Where This Crew Shines

Scenario 1: The Researcher Drowning in Papers

You're a PhD candidate with 200+ papers scattered across folders, half-remembered insights, and a supervisor meeting in 20 minutes. Traditional approach: panic-search, open 15 tabs, fail to find that crucial citation. With the Crew: "What did we decide about the pricing strategy?" The Seeker synthesizes across your vault with citations. "Process my meeting recording" triggers Transcriber → Architect chain when a new project emerges. Your weekly /vault-audit surfaces broken links and orphaned notes before they become problems.

Scenario 2: The Remote Worker With Email PTSD

Your Gmail has 4,327 unread messages. Important deadlines hide in promotional noise. You've missed three client meetings this quarter. With the Crew: "Check my email" triggers /email-triage, which scans, scores priority, saves critical emails as vault notes, and chains Sorter to file deadline-related items. The /deadline-radar skill builds a unified timeline. /meeting-prep generates comprehensive briefs before calendar events. The Postman doesn't just bridge email—it transforms it into actionable knowledge in your trusted system.

Scenario 3: The Non-Native English Speaker Excluded by Tools

Every "smart" productivity system forces you to think in English. Your ideas lose nuance; your workflow gains friction. With the Crew: "Salva questa nota veloce..." and Scribe responds in Italian. "Vérifie mon email..." and Postman replies in French. The entire system adapts to your cognitive language, not the other way around. This isn't translation layered on top—it's native multilingual operation at the agent level.

Scenario 4: The Person With Brain Fog Who Can't Maintain Systems

You've built beautiful Notion dashboards. Abandoned them. Tried PARA in Obsidian. Collapsed under maintenance load. The problem isn't you—it's that maintenance is a second job. With the Crew: You literally don't maintain anything. The Librarian runs weekly health checks. The Sorter files nightly. The Connector discovers links you'd never make. The /defrag skill runs 5-phase vault optimization automatically. You talk; they work. The system degrades gracefully because it's designed for people whose executive function is already compromised.

Step-by-Step Installation & Setup Guide

Prerequisites

• Obsidian (free version works)
• One supported AI platform: Claude Code, Gemini CLI, OpenCode, or Codex CLI
• Git installed
• Terminal/basic command line comfort (or a tech-savvy friend for 60 seconds)

Step 1: Create Your Obsidian Vault

Open Obsidian. Create a new vault, or identify an existing one you want to enhance. Note the full path to this vault folder.

Step 2: Clone the Repository

Open your terminal and navigate to your vault directory:

cd /path/to/your-vault
git clone https://github.com/gnekt/My-Brain-Is-Full-Crew.git

This creates a My-Brain-Is-Full-Crew/ directory inside your vault containing all agents, skills, scripts, and documentation.

Step 3: Run the Platform Installer

Navigate into the cloned directory and execute the launcher:

cd My-Brain-Is-Full-Crew
bash scripts/launchme.sh

The script interactively asks which platform to target. Choose from:

• claude-code → installs to .claude/, dispatcher as CLAUDE.md
• gemini-cli → installs to .gemini/, dispatcher as GEMINI.md
• opencode → installs to .opencode/, dispatcher as AGENTS.md
• codex-cli → installs to .codex/ + .agents/, dispatcher as AGENTS.md

For automated deployment, specify directly:

bash scripts/launchme.sh --platform claude-code

The installer builds platform-native configurations from the single source codebase, translating agents, skills, references, hooks, and MCP servers to each platform's format.

Windows users with Codex CLI: Run inside WSL (Windows Subsystem for Linux). Native Windows support is experimental.

Step 4: Initialize Your Vault

Open your AI platform inside your vault folder (this context-sensitivity ensures agents only activate for this project). Then simply say:

"Initialize my vault"

The /onboarding skill launches a guided conversation:

1. Identity capture: Name, preferred language, role, what brought you here
2. Needs assessment: Which agents to activate, life areas to manage
3. Integration setup: Gmail and Google Calendar connections (optional)

The Architect then generates Meta/vault-map.md (mapping your folder names to internal tokens), creates your folder structure, saves your profile, and leaves a welcome note. Existing vaults are preserved—the Architect scans and adapts, never forcing renames.

Step 5: Update Workflow

After git pull for new features:

cd /path/to/your-vault/My-Brain-Is-Full-Crew
git pull
bash scripts/updateme.sh

The updater auto-detects your installed platform. Specify explicitly if multiple:

bash scripts/updateme.sh --platform opencode

Only changed files update; your vault notes remain untouched.

REAL Code Examples From the Repository

Example 1: Installation Commands (From Quick Start)

The actual setup uses these exact commands from the README:

# Navigate to your Obsidian vault
cd /path/to/your-vault

# Clone the crew repository inside your vault
git clone https://github.com/gnekt/My-Brain-Is-Full-Crew.git

# Enter the repository directory
cd My-Brain-Is-Full-Crew

# Run the interactive installer
bash scripts/launchme.sh

What's happening here? The repository is designed to live inside your vault, not as a separate tool. This co-location is architecturally significant: agents read and write directly to your vault filesystem without API layers or sync delays. The launchme.sh script is a sophisticated build system that parses the platform-agnostic source files in agents/ and skills/, then compiles them to platform-specific formats. For Claude Code, this means generating .claude/agents/ with proper CLAUDE.md dispatcher syntax; for Codex CLI, it produces .toml agent configurations and TOML-based MCP server definitions.

Example 2: MCP Server Configuration for Apple Contacts

The repository includes this exact configuration for integrating with Apple Contacts:

{
  "mcpServers": {
    "apple-contacts": {
      "command": "npx",
      "args": ["-y", "@griches/apple-contacts-mcp"]
    }
  }
}

Platform-specific adaptation required: For Claude Code, Gemini CLI, and OpenCode, this JSON goes in your .mcp.json. But Codex CLI uses TOML instead—you'd add this to .codex/config.toml under [mcp_servers.apple-contacts]. This example reveals the project's thoughtful abstraction: the same capability surfaced through different configuration dialects. The @griches/apple-contacts-mcp package provides full CRUD access to Apple Contacts, enabling the /contact-sync skill to automatically sync contacts when you reply to emails or on demand. The -y flag auto-accepts the npx prompt, enabling unattended operation.

Example 3: Vault Structure (Hybrid PARA + Zettelkasten)

The default folder structure that the Architect creates during onboarding:

00-Inbox/          # Capture everything here first
01-Projects/       # Active projects with deadlines
02-Areas/          # Ongoing responsibilities
03-Resources/      # Reference material, guides, how-tos
04-Archive/        # Completed or historical content
05-People/         # Your personal CRM
06-Meetings/       # Timestamped meeting notes
07-Daily/          # Daily notes and journals
MOC/               # Maps of Content (thematic indexes)
Templates/         # Obsidian note templates
Meta/              # Vault config, agent logs, health reports

Critical design insight: The numbered prefix enforces explicit ordering in Obsidian's file explorer, creating a processing pipeline. Everything lands in 00-Inbox; the Sorter promotes items through the numbered stages. This isn't mere aesthetics—it's workflow as structure. The MOC/ (Maps of Content) directory implements Zettelkasten-style thematic indexes, while the numbered folders operationalize PARA (Projects, Areas, Resources, Archive). The Meta/ directory is where the crew's operational knowledge lives: vault-map.md translates your folder names to internal tokens, enabling the system to adapt to existing vaults rather than forcing renames.

Example 4: Post-Install Vault Layout (Platform-Dependent)

After running launchme.sh, your vault's actual filesystem looks like:

your-vault/
├── .claude/                    # Platform directory (example: Claude Code)
│   ├── agents/                 # Lightweight reactive agents
│   ├── skills/                 # Specialized multi-step skills
│   ├── hooks/                  # File protection and validation hooks
│   └── references/             # Shared documentation
├── Meta/
│   ├── vault-map.md            # Folder role mappings (auto-generated)
│   └── scripts/                # Orchestra scripts (permission-free commands)
├── CLAUDE.md                   # Dispatcher file (platform-specific name)
├── My-Brain-Is-Full-Crew/      # Source repo (for updates)
└── ... your Obsidian notes

Architecture exposed: The dispatcher file (CLAUDE.md, GEMINI.md, or AGENTS.md) is the entry point your AI platform reads when opened in this directory. It contains the routing logic that checks skills first, then falls through to agents. The orchestra/ scripts in Meta/scripts/ are permission-free operations—pre-approved commands agents can execute without user confirmation, enabling smooth chaining. The separation of .claude/ (platform config) from My-Brain-Is-Full-Crew/ (source for updates) means you can pull new versions without overwriting your platform-specific customizations.

Advanced Usage & Best Practices

Mobile Capture via Remote Control

Claude Code's Remote Control feature lets your phone act as a remote interface to your computer-running crew. Capture thoughts on walks, check email from couches, search your vault from supermarkets. Your computer hosts the actual agents; your phone is just the glass. Setup takes 2 minutes—see the mobile access guide.

Custom Agent Design Patterns

When creating custom agents, leverage the Architect's guided conversation but think in terms of trigger conditions and action boundaries. The best custom agents have narrow, well-defined responsibilities that complement rather than overlap with core crew members. The dispatcher auto-discovers them, so naming conventions matter—prefix with domain for clarity (budget-tracker, not tracker).

Integration Stacking

You can run both gws (Google Workspace CLI) and hey (Hey.com CLI) simultaneously if you manage multiple email identities. The Postman agent routes appropriately. For read-only fallback when full CLI tools aren't configured, MCP connectors auto-setup during launchme.sh provides limited but functional email and calendar access.

Defensive Vault Hygiene

While the Librarian handles automated maintenance, run /vault-audit manually before major life transitions (project completions, role changes). The 7-phase audit catches structural debt that gradual degradation hides. Similarly, use /tag-garden quarterly—tag proliferation is the silent killer of vault navigability.

Language Consistency Within Conversations

While the crew handles multilingual input, single conversations should stick to one language. Mixing languages in one session can confuse context maintenance, especially for skills with multi-step state. Start fresh if you need to switch languages.

Comparison With Alternatives

The verdict: My Brain Is Full Crew occupies a unique position—open-source, locally-hosted, multi-agent orchestration with deep Obsidian integration. It's not as polished as commercial alternatives, but it's the only option that combines true agent coordination with offline-first vault ownership and explicit wellness-aware design philosophy. For privacy-conscious users, researchers with sensitive data, or anyone who's watched cloud services disappear, the local-first architecture is non-negotiable.

FAQ: What Developers Actually Ask

Is this just a bunch of prompts?

No. While prompts are critical, the system includes a build toolchain (adapters/, scripts/launchme.sh), dispatcher logic for agent/skill routing, orchestra scripts for permission-free operations, and MCP server integrations. It's infrastructure, not just prompting.

How does this compare to building my own Obsidian + Claude setup?

You could. Many have. But you'll rebuild the dispatcher, the agent isolation, the skill state management, the platform abstraction layer, and the update mechanism. This project represents hundreds of hours of refinement by someone who actually lives with the system daily.

What about data privacy?

Your vault stays local. AI platform APIs process your queries, but your note contents reside on your filesystem. The creator explicitly designed this for personal use on your own data—GDPR/CCPA compliance is your responsibility if you process third-party emails.

Can I use this without Gmail?

Absolutely. Only the Postman agent and related skills require email integration. All knowledge management agents (Scribe, Seeker, Connector, Librarian, etc.) work with purely local vault operations.

What if an agent makes a mistake?

Conservative defaults: agents archive, never delete. They ask before major changes. The hooks/ system provides file protection and validation. And git versioning your vault provides ultimate recovery capability.

How do I contribute or report issues?

The creator actively solicits contributions—see CONTRIBUTING.md. Every PR welcome, from prompt improvements to new agents to bug reports. The project explicitly states: "I won't be precious about my code."

Will this work with my existing vault?

Yes. The Architect scans existing folders during onboarding and creates Meta/vault-map.md to map your structure to internal tokens. No forced renames, no migration trauma.

Conclusion: Your Brain Deserves a Crew, Not Another Tool

Here's the uncomfortable truth: productivity tools fail because they add work. Every system that requires maintenance, organization, or ritualistic upkeep eventually collapses under the weight of the very overwhelm it promised to solve.

My Brain Is Full Crew inverts this entirely. It's not a tool you operate; it's a team you delegate to. The chat interface eliminates cognitive switching costs. The multi-agent architecture handles complexity without overwhelming single-context windows. The multilingual support respects your cognitive native language. The wellness-aware philosophy acknowledges that knowledge work is embodied, not abstract.

For researchers drowning in papers. For developers with brain fog. For non-native English speakers excluded by Anglo-centric tooling. For anyone who's ever abandoned a beautiful organizational system because maintaining it became a second job.

This is the system built by someone who actually experienced cognitive failure and engineered through it—not as optimization, but as survival.

Clone it. Initialize it. Talk to it. Your vault will organize itself while you sleep, while you walk, while you finally have the mental space to think about what matters instead of where you filed it.

The best organizational system is the one you actually use. This one, finally, uses itself.

⭐ Star the repository if it helps you—or if you believe overwhelmed people deserve better tools. Join the Discord community for support, custom agent sharing, and honest feedback about what works and what doesn't.

Your brain is full. Let the crew handle the overflow.