1Write the epic and split it
Write the epic as one note in the project’s kanban folder and keep it in the Note column, where no agent can claim it. It holds the goal, the constraints, and what done looks like for the whole thing. It never moves. Every ticket points back to it.
Then ask your agent to split it. You get a list of tickets, each with a title, acceptance criteria and its dependencies, named in plain words. Mo relates the tickets to the epic through topics and through the words they share. You do not wire links by hand.
2Let agents claim tickets
The agent lists the Todo column, picks the top ticket and calls mo_take_task with claim: true. The ticket moves to Doing in the same call that returns its work packet. If a second agent reaches for the same card, its claim comes back claimed: false and it takes the next one.
Progress goes on the ticket as comments, not in the chat. “Schema done, replay pending” on the card is worth more than the same sentence in a transcript. The next session, the reviewer and Ask Mo can all read it.
3Answer questions on the ticket
An agent that hits a decision it cannot make calls mo_request_human. That is a durable comment on the ticket or a card in the backlog, not a chat that blocks. Your agent cannot ping you, so it will also tell you in its own chat to check Morion. You answer when you see it, in the app or from any agent. The answer becomes part of the ticket’s story, and the next run reads it.
4Review, re-open, close
Done is your column. Agents stop at Review. When a reviewed ticket is not right, do not delete it and do not write a new one. Move it to Re-opened and say what is wrong in a comment. The card records who sent it back; in the card drawer, pick what happens next: New session, Ask human, or Resume session. Both travel with the card, so whoever picks it up next starts from the failure instead of from zero.
Move to Done yourself. Done is the one column only you move.
5Ask Mo what is left
Ask Mo, scoped to the project and to the open statuses. The answer lists what is in each column, with a citation per claim. It also lists the gaps: things the epic mentions that no ticket covers. That is the planning meeting, and it takes a minute.
6Questions
No. The claim is atomic: one agent gets the card and moves it to Doing, the other gets claimed: false and takes the next one.
It posts the question on the ticket, marked as awaiting a human. You answer in a comment, in the app or from any agent, and the next run reads it.
7What you have now
An epic that outlives any one session. The plan sits on the board, each ticket carries its own progress, questions and reasons, and any agent you point at the folder picks up the next unclaimed piece with the related notes and tickets in its packet. You move cards to Done and ask Mo for the state of play. Nothing depends on a transcript you have to find again.
Choose who works the board.
Give the tickets a knowledge layer.