That's it. No new tools, no new dependencies. Just three small additions to the same plain-markdown shape. Additive — v1.9.x folders gain new directories without losing anything.

Get going now

1. Clone or download the repo into a folder you'll actually use.
2. Open the folder in your LLM tool (Claude Code, Codex CLI, Gemini CLI, Cursor, or Obsidian + chat plugin).
3. Paste the contents of initialize yourself in this folder as your first message.
4. The LLM reads ADAPTER-PROMPT.md, writes a tool-specific pointer file (CLAUDE.md, GEMINI.md, etc.), and reports the team is online.
5. Ask "Who are you?" and you'll see Larry is at your service.
6. Ask "What's open?" and Larry walks the new Team Knowledge/tasks/open/ folder for you.

That's the whole setup. There is no install step.

Once you have the team online and start using it, you'll hit moments where you wonder why a folder is shaped the way it is. That's where the courses come in. They teach the WHY behind every folder choice, so you understand why the team is built like real humans operating on the ICOR methodology, not just another AI scaffold. More on that below.

What you get

A working knowledge system, fully assembled, that does this on day one:

• Organizes your life from a single daily journal. You write what happened. The team files the people, projects, decisions, and ideas where they belong. Connections between notes are made for you.
• Remembers unfinished work for you. When something can't finish in one turn, the team writes it down as a task in Team Knowledge/tasks/. Next session, Larry walks tasks/open/ and surfaces what's waiting. The team genuinely picks up where it left off.
• Carries learning forward. Each specialist keeps a journal of durable insights at Team/<Name>/journal/. When a task references one of their entries, they re-read their own past thinking before starting work.
• Runs in any LLM you already use. Claude Code (and Claude Cowork), Codex CLI, Gemini CLI, Cursor, ChatGPT, Obsidian with a chat plugin. The same scaffold, the same team, the same files. You change models. Your knowledge doesn't move. Session-log triggers (close session, keep this in mind, let's realign, etc.) work with any LLM that reads AGENTS.md — ChatGPT, Claude, Gemini, Cursor, Cline, Codex, and the rest. Not Claude-only.
• Stays in plain markdown. Every note is a .md file. You can read it without the AI. You can grep it. You can sync it with Dropbox or git. You can open it in Obsidian and keep working with no AI at all.
• Upgrades to SQLite when you outgrow plain files. Once your myPKA gets large, paste the prompt at Team Knowledge/SOPs/SOP-002-convert-mypka-to-sqlite.md into your LLM. Markdown stays canonical. SQLite becomes a fast lookup layer on top.
• Imports from any PKM tool. Drop your existing notes (Heptabase backup, Notion export, Obsidian vault, Roam graph, Logseq folder, Apple Notes export, Evernote dump, Tana via MCP, etc.) anywhere on disk, then ask your LLM something like "import my Notion export from ~/Downloads/notion.zip" or "how do I bring in my Heptabase backup?". The scaffold ships with [[Team Knowledge/Workstreams/WS-002-import-external-knowledge-base]] — the LLM follows it to extract entities (people, organizations, projects, goals, habits, topics, key elements, documents), normalize wikilinks to the slug form, copy attachments into PKM/Images/YYYY/MM/, and place files into the right folders, asking clarifying questions where needed and never overwriting your myPKA without explicit approval. Works with any LLM that reads AGENTS.md.

There is no lock-in. The whole system is text on your disk. It works in Obsidian today. It upgrades to SQLite when you want it. It runs on whichever model or LLM you prefer, and it keeps running when you switch.

Who this is for

• Knowledge workers who want a local folder setup instead of handing over their knowledge to Notion, Tana etc. and want an AI team that actually files things. Transparent and accessible.
• Founders and operators running multiple projects who need a knowledge system that thinks across People, Topics, Goals, Habits, and Key Elements without manual cross-linking.
• Parents and generalists with too many inputs (school stuff, health stuff, ideas, contacts) and no structure to hold it.
• AI tinkerers who want a real reference architecture for a multi-agent setup, not a toy demo.

If you've ever opened a blank Obsidian vault and didn't know where to put anything, this is for you. (Yes — myPKA is fully Obsidian-compatible. Open the folder as an Obsidian vault and everything just works.)

Meet the team

Nine specialists ship pre-loaded. You only ever talk to Larry. Larry routes.