# My Brain Is Full Crew: an agent team for Obsidian, nutrition and mental health > My Brain Is Full Crew is a Shell-based collection of 8+ agents and 14 skills that runs on Claude Code, Gemini CLI, OpenCode or Codex CLI and organises an Obsidian vault, including notes about food and mental state. It is aimed at people whose working memory is already full, and it says so itself. **gnekt/My-Brain-Is-Full-Crew** — Built by a PhD whose memory was failing, whose diet was a mess, and whose anxiety had its own agenda. Most second brain tools ignore the fact that your brain doesn't work in isolation: your body and your mental health are part of the system too. This crew handles all three: knowledge, nutrition, and mental wellness. - Repository: https://github.com/gnekt/My-Brain-Is-Full-Crew - Stars: 3,813 · Forks: 370 - Language: Shell - License: NOASSERTION - Published: 2026-09-23 · Updated: 2026-09-23 · Language: en - Canonical page: https://hysenlabs.com/projects/gnekt-my-brain-is-full-crew ## The problem My Brain Is Full Crew is aimed at The README opens with an origin story rather than a feature list. The author describes being a PhD researcher whose memory "started slipping", not as a diagnosis but as a budget running out: forgetting what was read, losing track of conversations, feeling behind. The stated gap is that existing Obsidian plus Claude setups are "clever note-capture tools, glorified search engines for your second brain". That is a fair description of the category, and it is the reason this project exists. The intended audience is narrow and stated plainly: PhD students and researchers, people with brain fog or an overloaded working memory, non-native English speakers, and anyone who tried Obsidian and quit because maintaining it felt like a second job. The distinguishing claim is that the system covers nutrition and mental wellness alongside knowledge, on the argument that a brain does not work in isolation from the body. Whether that is a feature or a category error depends on what you want from a note tool, and the README does not try to settle it. ## How the crew, the dispatcher and the skills fit together The repository is not a single program. The top level holds agents/, skills/, hooks/, mcp/, orchestra/, adapters/, references/ and scripts/, plus a DISPATCHER.md and a settings.json. The README describes 8+ agents and 14 skills, and names six of the agents: Architect (vault structure and setup), Scribe (text capture), Sorter (inbox triage), Seeker (search with citations), Connector (knowledge graph links) and Librarian (weekly vault maintenance). Each agent has a role and a stated superpower rather than a function signature. The coordination mechanism is the dispatcher. The README gives one concrete example: when the transcription agent processes a meeting and finds a new project, the dispatcher chains the Architect to create the folder structure. That is the whole architectural idea in miniature. Agents are not a menu you pick from; one agent's output can trigger another. The adapters/ directory is what lets the same agent definitions run under four different CLIs, which is why the README can claim "One codebase. Four platforms. Same crew." The interface is the chat window. The README states that the author does not browse Obsidian, drag files, or maintain folder structures by hand, and that agents match the language the user writes in, listing Italian, French, German, Spanish and Japanese as examples. That language-matching claim is not backed by a mechanism in the README, so treat it as a design intention rather than something you can verify from the text. ## Installing it and getting one note filed The README does not contain an install command, a package name, a version or a port. What it does document is the shape of the first run: you open one of the four supported CLIs, and the Architect runs onboarding, which includes asking you to accept the terms before proceeding. The repository ships a .claude-plugin/ directory and a settings.json, which is where a Claude Code user would expect plugin configuration to live, but the README does not spell out the steps. So the honest first move is to read the two documents the README points at, TERMS_OF_USE.md and docs/DISCLAIMERS.md, then clone the repository and open it in your CLI of choice. Until the README gains an install section, do not expect a one-line bootstrap. Once onboarding is done, the first real use is a brain dump. The Scribe agent is described as transforming "messy, typo-filled, stream-of-consciousness dumps into clean notes", and the Sorter is described as emptying the inbox every evening and routing each note to its home. A first session therefore looks like this: write whatever is in your head, in your own language, and let the crew file it. What you should see is a cleaned note in the vault and, if the dispatcher fires, follow-on structure created by the Architect. The custom agent path is the other thing worth trying early. The README says you say "create a new agent" and the Architect walks you through a conversation to design one, with no code or config files to edit. The README's own examples are budget-tracker, wardrobe-coach, home-inventory, project-pulse and client-tracker. It also states that custom agents are your responsibility and carry no warranty. ## What the README does not promise There are no releases listed for this repository, so there is no changelog to read and no version to pin. The last push was on 2026-06-27, which is under six months before today, so calling it maintained is defensible on that fact alone, but a single push date is thin evidence for anyone planning to run this over a vault they care about. The licence field is NOASSERTION even though the README badge says MIT and a LICENSE file sits at the top level. That mismatch matters: if you need to know the terms before you redistribute or embed this in something else, read LICENSE directly rather than trusting the badge. The bigger limitation is behavioural. The system acts on your vault automatically. The Sorter routes notes, the Librarian does weekly maintenance, and the dispatcher chains agents without asking. The README's own disclaimers say the software is provided as is, that you should back up your vault, and that the author accepts no liability. There is no documented dry-run mode, no rollback command, and no undo described anywhere in the README. If you are the kind of person who wants to review every file move, this design will fight you. The nutrition and mental wellness coverage is the other place to be careful. The README presents it as part of the system, not as medical or dietary advice, and the full disclaimers live in docs/DISCLAIMERS.md. If you want a tool that only stores notes and never touches the subject matter, the framing here will feel like scope creep. ## Where it sits next to plain Obsidian and generic capture tools The obvious alternative is Obsidian plus a single capture plugin, or one of the general Obsidian plus Claude setups the README describes as glorified search engines. The difference in approach is coordination. A capture plugin waits for you to invoke it and does one thing. My Brain Is Full Crew runs several agents with defined roles, and the dispatcher lets one agent's finding change what another agent does. That is more machinery, and more machinery means more places for a wrong routing decision to land in your vault. The second alternative is a hosted second-brain product. Those give you a support channel and a release cadence. This project gives you a Shell codebase you can read, an adapters/ directory that targets four CLIs, and no releases. The trade is control against support, and the README does not pretend otherwise. The third comparison is doing nothing and keeping a flat folder of notes. For someone with a small vault and no overload problem, that is genuinely better: no onboarding, no terms acceptance, no agents moving files. The project is explicit that it is for people who are drowning, not for people who want to optimise a system that already works. ## Maintenance cost, licence and the terms you accept at onboarding The upgrade story is the weakest part of the picture. With no releases, there is nothing to upgrade between; you track the main branch, and the last push was on 2026-06-27. If you fork the crew to add custom agents, you own the merge work for every change upstream makes to the agent definitions and the dispatcher. The runtime cost is different from the install cost. Agents that transcribe, search and synthesise across a vault consume model calls through whichever CLI you chose, and the README gives no token, latency or cost figures. Anyone budgeting for this should assume the cost scales with how much you dump into the vault, not with the size of the repository. On licensing, the README badge says MIT, the LICENSE file is present, and the platform metadata reports NOASSERTION. Those do not agree, and the README's own disclaimers push compliance onto you: it states that the software is for personal use on your own data, and that you are responsible for GDPR or CCPA compliance if you process third-party data such as emails containing other people's information. That is a real constraint for anyone pointing the email triage agent at a shared inbox. Read LICENSE and TERMS_OF_USE.md together before you decide, and treat the onboarding acceptance step as the moment that decision becomes binding. ## Conclusion Adopt My Brain Is Full Crew if you already live in Obsidian and one of the four supported CLIs, and you want capture, triage and search to happen from a chat window instead of a folder tree. Do not adopt it if you need a supported product with a release history, or if you expect the nutrition and mental wellness side to be clinical: the README frames those as part of a personal system, not as advice, and it disclaims warranty. Verify three things before you commit your vault: that your CLI is one of Claude Code, Gemini CLI, OpenCode or Codex CLI, that you have read TERMS_OF_USE.md and docs/DISCLAIMERS.md, and that you have a backup you can restore, because the README says the software is provided as is. ## FAQ ### What is My Brain Is Full Crew? It is a Shell codebase containing 8+ AI agents and 14 skills that manage an Obsidian vault from a chat interface, covering knowledge capture, nutrition notes and mental wellness notes. Its README describes it as built for people whose working memory is already overloaded rather than for people optimising an organised system. ### Which platforms does My Brain Is Full Crew run on? The README lists Claude Code, Gemini CLI, OpenCode and Codex CLI, and the repository has an adapters/ directory for them. The README summarises this as one codebase, four platforms, same crew. ### Does My Brain Is Full Crew work in languages other than English? The README states that the system works in any language and that the agents match the language you write in, naming Italian, French, German, Spanish and Japanese as examples. It does not describe the mechanism behind the language matching. ### Can I build my own agents in My Brain Is Full Crew? Yes. The README says you say "create a new agent" and the Architect walks you through a conversation to design one, with no code or config files to edit. It also states that custom agents are created by you, run on your data, and carry no warranty from the project. ### What licence does My Brain Is Full Crew use? The README badge says MIT and a LICENSE file is present at the top level, but the repository metadata reports NOASSERTION. Read LICENSE directly if the terms matter to you. ## Sources - [gnekt/My-Brain-Is-Full-Crew on GitHub](https://github.com/gnekt/My-Brain-Is-Full-Crew) - [Issues](https://github.com/gnekt/My-Brain-Is-Full-Crew/issues) - [README](https://github.com/gnekt/My-Brain-Is-Full-Crew/blob/main/README.md) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/gnekt-my-brain-is-full-crew