A team of 8+ AI agents and 14 specialized skills that manage your Obsidian vault
so your brain doesn't have to.

You talk. They organize, file, connect, search, transcribe, and triage your email.
In any language. On the platform you already use.

One codebase. Four platforms. Same crew.

I'm a PhD researcher. I've spent years training my brain to hold enormous amounts of information: papers, ideas, deadlines, people, half-baked theories at 2am. And for a while, it worked.

Then it didn't.

Memory started slipping. Not dramatically (no diagnosis, no crisis) just the slow, creeping realization that the mental budget was getting empty, and things were falling through the cracks. I'd forget what I'd read. Lose track of conversations. Feel constantly behind, constantly overwhelmed.

I started looking for solutions. I found a lot of Obsidian + Claude setups online. They were mostly clever note-capture tools, glorified search engines for your second brain. Useful. But not what I needed.

What I needed wasn't just a memory extension. I needed a brain dump system, something that could help me organize not just my knowledge, but my life: my overwhelmed mind, my wrecked physical health, the avalanche of emails and commitments and things I should have done last week.

So I built this.

Most "AI + Obsidian" tools are built for people who already have their life together and want to optimize. This one is for people who are drowning and need a lifeline.

1. The chat IS the interface. I don't browse Obsidian. I don't drag files around. I don't maintain complex folder structures manually. I just talk to Claude. Everything else happens automatically.

2. It speaks your language, literally. The system works in any language. You shouldn't need to think in English to manage your brain. Just talk in Italian, French, German, Spanish, Japanese, whatever feels natural. The agents match you.

3. The agents coordinate through a dispatcher. When the transcription agent processes a meeting and discovers a new project, the dispatcher automatically chains the Architect to create the folder structure. It's a crew, not a collection of isolated tools.

4. 8 agents are just the starting point. Build your own. The crew ships with 8 agents. But your life isn't generic, and your system shouldn't be either. Say "create a new agent" and the Architect walks you through a conversation to design one from scratch. No code, no config files, no templates to edit. You describe what you need, it builds it.

Custom agents coordinate with the core crew, get discovered automatically by your agent platform, and respond in your language. They just solve the problems that are specific to your life.

Your custom agents, your responsibility. Custom agents are created by you and run on your data. The project provides no warranty on their behavior. See Terms of Use.

• PhD students, researchers, academics drowning in papers and commitments
• Anyone with brain fog, or just an overloaded working memory
• Non-native English speakers who want a system that works in their language
• Anyone who's tried Obsidian before and gave up because it felt like a second job

If you've ever thought "I need to get organized, but I'm too exhausted to get organized", this is for you.

Please read the full disclaimers before using this project.

Key points:

• This software is for personal use on your own data. You are responsible for GDPR/CCPA compliance if you process third-party data (e.g., emails containing other people's information).
• No warranty. Provided "as is". Back up your vault. The author accepts no liability.
• No responsibility for forks or misuse. This is a personal productivity tool. Malicious repurposing is explicitly condemned.

By using this software, you agree to the Terms of Use. During onboarding, the Architect will ask you to explicitly accept these terms before proceeding.

Agents + Skills = the full system. Each agent handles quick, reactive tasks. For complex multi-step workflows (like onboarding, email triage, or vault audits), the dispatcher routes to one of 14 specialized skills that run as guided conversations. See the Skills section below.

Skills handle the complex, multi-step workflows that need conversational context. While agents are great for quick, one-shot tasks, some operations — like onboarding or email triage — require a back-and-forth conversation. Skills run in the main conversation context, so they can ask questions, wait for answers, and maintain state naturally.

The dispatcher automatically routes your message to the right skill or agent. You don't need to remember which is which.

You talk naturally  →  Dispatcher checks skills first  →  If match: invokes skill
                                                         →  If no match: invokes agent  →  Your vault gets updated

The dispatcher has two delegation mechanisms. Skills handle complex, multi-step conversational flows (onboarding, email triage, vault audits). Agents handle quick, reactive single-shot operations (capture a note, search the vault, create a folder). Skills are checked first because they cover the most involved workflows. If no skill matches, the dispatcher falls through to agents.

Each crew member is an isolated AI with its own system prompt, tool restrictions, and model assignment. You clone the repo into your vault, run a setup script, and from that moment on you manage everything through conversation. No GUI, no drag-and-drop, no manual file management.

graph TB
    User((You))
    Dispatcher["Dispatcher"]

    User -->|"talk naturally"| Dispatcher
    Dispatcher -->|"skill match?\ninvoke skill"| Skills
    Dispatcher -->|"no skill match?\ninvoke agent"| Agents
    Dispatcher -->|"chains agents when needed"| Agents

    subgraph Skills["Specialized Skills (14)"]
        direction TB
        Onboarding["/onboarding"]
        EmailTriage["/email-triage"]
        Transcribe["/transcribe"]
        InboxTriage["/inbox-triage"]
        VaultAudit["/vault-audit"]
        MoreSkills["... +9 more"]
    end

    subgraph Agents["The Crew (8 agents)"]
        direction TB

        subgraph Core["Core: Knowledge Management"]
            Architect["Architect\nVault Setup"]
            Scribe["Scribe\nNote Capture"]
            Sorter["Sorter\nInbox Triage"]
            Seeker["Seeker\nSearch"]
            Connector["Connector\nKnowledge Graph"]
            Librarian["Librarian\nVault Health"]
        end

        subgraph External["Integrations"]
            Transcriber["Transcriber\nAudio & Meetings"]
            Postman["Postman\nEmail & Calendar"]
        end
    end

    Skills <-->|"read & write"| Vault
    Agents <-->|"read & write"| Vault

    subgraph Vault["Your Obsidian Vault"]
        direction LR
        Inbox["00-Inbox"]
        Projects["01-Projects"]
        Areas["02-Areas"]
        Resources["03-Resources"]
        Archive["04-Archive"]
        Daily["07-Daily"]
    end

    style User fill:#7c3aed,stroke:#5b21b6,color:#fff
    style Dispatcher fill:#3b82f6,stroke:#2563eb,color:#fff
    style Skills fill:#fef3c7,stroke:#f59e0b
    style Core fill:#e0e7ff,stroke:#818cf8
    style External fill:#dbeafe,stroke:#60a5fa

sequenceDiagram
    participant U as You
    participant D as Dispatcher
    participant TS as /transcribe skill
    participant A as Architect agent
    participant ES as /email-triage skill
    participant S as Sorter agent

    U->>D: "Process my meeting recording"
    D->>TS: invokes /transcribe skill
    TS->>TS: multi-step conversation:<br/>asks questions, transcribes,<br/>creates structured note
    TS-->>D: "Suggested next agent: Architect<br/>(new project mentioned)"
    D->>A: chains Architect agent
    A->>A: creates folder structure

    U->>D: "Check my email"
    D->>ES: invokes /email-triage skill
    ES->>ES: scans Gmail, scores priority,<br/>saves important emails as notes
    ES-->>D: "Suggested next agent: Sorter<br/>(deadline notes in Inbox)"
    D->>S: chains Sorter agent
    S->>S: files notes to correct locations

The Crew works on multiple agent platforms. The installer builds from a single source and deploys to your platform of choice:

If you omit --platform, the installer asks you to choose. Each platform gets agents, skills, references, hooks, and MCP servers translated to its native format. launchme.sh installs everything automatically.

Codex CLI on Windows: Codex CLI's Windows support is experimental. Running inside WSL (Windows Subsystem for Linux) is strongly recommended. See docs/codex-cli.md for details.

Your vault follows a hybrid PARA + Zettelkasten structure. These are the default folder names — if you already have a vault with different names, the Crew adapts to yours during onboarding (see Vault Mapping):

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

Prerequisites: Obsidian (free) and one of the supported agentic platforms.

Open Obsidian and create a new vault (or use an existing one).

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

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

The script asks you to pick a platform, then builds and installs the agents and skills into your vault. When your agent platform is open in your vault folder, the agents activate automatically. When you're in any other project, they don't.

Never used a terminal before? See the step-by-step guide for beginners. It walks you through everything, or just show this page to a tech-savvy friend. It takes 60 seconds.

Open your agent platform inside your vault folder and say:

"Initialize my vault"

The /onboarding skill starts a friendly guided conversation:

1. Who are you? Name, language, role, what brought you here
2. What do you need? Which agents to activate, which areas of life to manage
3. Integrations Gmail and Google Calendar connections

After onboarding, the Architect generates Meta/vault-map.md (mapping your folder names to the Crew's internal tokens), creates your entire vault folder structure, saves your profile, leaves you a welcome note, and you're ready to go. If you already have an existing vault, the Architect scans your folders and adapts — no renaming needed.

The Crew is built in English but responds in whatever language you write in. Italian, French, Spanish, German, Portuguese, Japanese: just talk, and the agents match you.

"Salva questa nota veloce..."          → Scribe responds in Italian
"Vérifie mon email..."                 → Postman responds in French
"Was habe ich diese Woche geplant?"    → Seeker responds in German
"Check my inbox"                       → Sorter responds in English

No translations to install. No language packs. It just works.

If you use Claude Code, you can control the Crew from your phone using its Remote Control feature. Your computer runs Claude Code locally (with full vault and agent access), and your phone acts as a remote interface through the browser or the Claude mobile app.

Capture a quick thought on a walk. Check your email from the couch. Search your vault from the supermarket. Everything runs on your computer; your phone is just the remote.

Full setup guide (takes 2 minutes)

Agents coordinate through a dispatcher-driven orchestration system. When an agent or skill finishes its task and detects work for another agent, it signals the dispatcher via a ### Suggested next agent section in its output. The dispatcher reads this and automatically chains the next agent:

• The /transcribe skill processes a meeting that introduces a new project -- the dispatcher chains the Architect to create the folder structure
• The /email-triage skill finds emails about deadlines -- the dispatcher chains the Sorter to file them
• The Connector finds orphan notes -- the dispatcher chains the Librarian to investigate
• The Sorter finds notes that belong to a new area -- the dispatcher chains the Architect to build it

No agent works in isolation. The crew is greater than the sum of its parts.

The Postman agent (and its related skills: /email-triage, /meeting-prep, /weekly-agenda, /deadline-radar) requires one of:

• Google Workspace CLI (gws) — full read/write access to Gmail and Google Calendar: search, read, archive, delete, label, send emails; create/update/delete calendar events. See docs/gws-setup-guide.md for setup.
• Hey CLI (hey) — for Hey.com accounts. Read/reply/compose emails, leverages Hey's pre-sorted mailboxes (Imbox, Feed, Paper Trail, Reply Later, Set Aside, Bubble Up). Calendar operations still use gws. See Hey CLI for installation.
• MCP connectors (read-only fallback) — launchme.sh sets up MCP servers automatically (format varies by platform). Limited to reading emails and calendar events, plus draft creation.

You can use gws and hey simultaneously if you have both Gmail and Hey.com accounts.

The /contact-sync skill syncs contacts to Apple Contacts on macOS. It requires the apple-contacts MCP server:

• @griches/apple-contacts-mcp — an MCP server that provides full CRUD access to Apple Contacts (search, create, update, delete contacts and groups).

Claude Code / Gemini CLI / OpenCode — add it to your .mcp.json inside the mcpServers object:

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

Codex CLI — add it to .codex/config.toml instead (Codex uses TOML, not .mcp.json). Follow the instructions in docs/codex-cli.md for the correct [mcp_servers.apple-contacts] table format.

Once connected, the /contact-sync skill auto-syncs contacts when you reply to emails or on demand ("sync contact", "add to contacts").

All other agents and skills work with just your local Obsidian vault. No integrations needed.

After pulling new changes from the repo:

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

The updater automatically detects which platform is installed in your vault (by checking for a platform-specific folder). If you have multiple platforms installed, it asks you to choose which one to update. You can also specify explicitly with --platform:

bash scripts/updateme.sh --platform opencode

Only changed files are updated. Your vault notes are never touched.

Essential: Templater, Dataview, Calendar, Tasks

Recommended: QuickAdd, Folder Notes, Tag Wrangler, Natural Language Dates, Periodic Notes, Omnisearch, Linter

My-Brain-Is-Full-Crew/               ← cloned inside your vault
├── agents/                          The 8 core agents
│   ├── architect.md                   Vault setup & onboarding
│   ├── scribe.md                      Text capture & note creation
│   ├── sorter.md                      Inbox triage & filing
│   ├── seeker.md                      Search & knowledge retrieval
│   ├── connector.md                   Knowledge graph & link analysis
│   ├── librarian.md                   Vault health & maintenance
│   ├── transcriber.md                 Audio & meeting transcription
│   └── postman.md                     Email & calendar integration
├── skills/                          The 14 specialized skills
│   ├── onboarding/SKILL.md            Full vault setup conversation
│   ├── create-agent/SKILL.md          Design a custom agent step by step
│   ├── manage-agent/SKILL.md          Edit, remove, or list custom agents
│   ├── defrag/SKILL.md                Weekly vault defragmentation
│   ├── email-triage/SKILL.md          Scan and prioritize unread emails
│   ├── meeting-prep/SKILL.md          Comprehensive meeting brief
│   ├── weekly-agenda/SKILL.md         Day-by-day week overview
│   ├── deadline-radar/SKILL.md        Unified deadline timeline
│   ├── transcribe/SKILL.md            Process recordings into structured notes
│   ├── vault-audit/SKILL.md           Full 7-phase vault audit
│   ├── deep-clean/SKILL.md            Extended vault cleanup
│   ├── tag-garden/SKILL.md            Tag analysis and cleanup
│   ├── inbox-triage/SKILL.md          Process and route inbox notes
│   └── contact-sync/SKILL.md          Sync contacts to Apple Contacts
├── orchestra/                       Named scripts for permission-free agent operations
│   ├── hey-imbox, hey-feed, ...       Hey mailbox wrappers
│   ├── tracker-today, tracker-search  Local tracker queries
│   ├── contact-lookup                 Person-based email search
│   └── vault-stats, vault-inbox       Vault health helpers
├── references/                      Shared agent documentation
├── scripts/
│   ├── launchme.sh                    First-time installer
│   └── updateme.sh                    Post-pull updater
├── docs/                            User-facing documentation
│   ├── getting-started.md             Step-by-step setup guide
│   ├── examples.md                    Real-world usage examples
│   ├── vault-mapping.md               Vault path tokenization guide
│   └── agents/                        Deep-dive into each agent
├── adapters/                        Platform adapters (build system)
│   ├── lib.sh                         Shared parsing and rewrite helpers
│   ├── claude-code/                   Claude Code adapter
│   ├── gemini-cli/                    Gemini CLI adapter
│   └── opencode/                      OpenCode adapter
├── mcp/servers.yaml                 MCP server definitions (source of truth)
├── LICENSE
├── README.md                        You are here
└── CONTRIBUTING.md

After running launchme.sh, your vault looks like (paths vary by platform):

your-vault/
├── .<platform>/              ← .claude/, .gemini/, .opencode/, etc.
│   ├── agents/               ← lightweight reactive agents
│   ├── skills/               ← specialized multi-step skills
│   ├── hooks/                ← file protection and validation hooks
│   └── references/           ← shared docs
├── Meta/
│   ├── vault-map.md          ← maps folder roles to your actual paths (created during onboarding)
│   └── scripts/              ← orchestra scripts (permission-free agent commands)
├── CLAUDE.md / GEMINI.md / AGENTS.md / ...  ← dispatcher (platform-specific name)
├── My-Brain-Is-Full-Crew/    ← the repo (for updates)
└── ... your Obsidian notes

Codex CLI uses a split layout instead of a single platform directory:

your-vault/
├── .codex/
│   ├── agents/               ← 8 core agents (.toml format)
│   ├── references/           ← shared docs
│   └── config.toml           ← MCP servers + profiles + sandbox policy
├── .agents/
│   └── skills/               ← 14 specialized skills
├── Meta/
│   └── scripts/              ← orchestra scripts
├── AGENTS.md                 ← dispatcher (Codex reads this)
├── My-Brain-Is-Full-Crew/    ← the repo (for updates)
└── ... your Obsidian notes

This started as one person's survival tool. I'm sharing it because I think it can help others, but I know it can be much better, and I need help from people who know prompt engineering, agentic platforms, and Obsidian better than I do.

Every single PR is welcome. I mean it. If you see something that could be improved (a better prompt structure, a smarter agent behavior, a more elegant architecture) please submit it. I won't be precious about my code. The goal is to help people, not to protect my ego.

If you want to:

• Improve an agent or skill: make it smarter, add a mode, fix edge cases
• Fix my prompts: if you know better patterns, teach me
• Propose a new crew member: a new agent or skill for a new domain
• Report a bug: something an agent does wrong
• Add examples: share how you use the Crew
• Just tell me what I'm doing wrong: I'll listen

...PRs, issues, and honest feedback are all welcome. See CONTRIBUTING.md.

• Update the installer to auto-install the orchestra scripts (currently a manual copy-paste step)
• Completely decouple the installer from the platform (currently has some platform-specific logic that could be moved to the adapters or to some kind of config file)

"The best organizational system is the one you actually use."

The Crew is designed for people who are overwhelmed, not for people who enjoy organizing. Every design decision prioritizes minimum friction:

• Chat is the interface: no manual file management
• Skills handle the heavy lifting: multi-step workflows run as guided conversations
• Agents handle the quick stuff: filing, linking, capturing, searching
• Any language, any time: your brain shouldn't have to switch languages to stay organized
• Conservative by default: agents never delete, always archive. They ask before making big decisions.

If the Crew helps you, or if you just think it's a cool idea, consider starring this repo. It helps others find it, and it motivates continued development.

MIT: use it, modify it, share it. Just keep the attribution.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED. The authors are not liable for any claim, damages, or other liability arising from the use of this software. See the MIT License for full terms.

Built by someone who got tired of forgetting things.

Get Started · Examples · Codex CLI Guide · Migrate to Codex · Meet the Agents · Contribute