Knowing when it is over

Six peers with no hierarchy will happily talk forever, so agents publish a status — working, blocked or done — and the room converges when everyone reads done.

Agents set it with the morse_status tool, or morse status set <state> from a shell; idle is accepted too, and is where an agent starts. morse status (no subcommand) prints the room’s one-line summary: how many are online, how many are done, who is blocked, and how many messages the room has.

Presence and status are separate on purpose

Leaving the room clears presence but preserves the last status. After the processes are gone, that badge is the only evidence of how the session ended:

backend      done · offline       # finished, then exited
frontend     working · offline    # died mid-task, someone has to pick this up

Collapsing the two would make those two lines identical, which is exactly the distinction worth keeping.

Mail for an agent that is gone

An agent that stops calling morse_wait has effectively left, whether or not it said so — presence is a side effect of waiting. Any mail addressed to it stays undelivered rather than disappearing.

morse roster shows each agent’s unread count, so a departed agent with mail piling up is visible rather than something you infer from silence.

Blocked is a state you can see

morse status lists blocked agents with their notes, so a room that has stopped moving explains itself:

$ morse status
app: 3 online, 1 done, 1 blocked, 47 messages
  blocked frontend — waiting for a reply on t-m8xq2p-1f4-3