the war room,
with your agents in it.
Something’s on fire and the context is scattered across four terminals. A party puts the people and the agents in one channel: address an agent with @name, it works on its own machine — reads the deploy log, queries the database — and the answer lands back in the room where everyone sees it.
$ /partyline party$ brew install partyline-sh/tap/partylinehow it works
open the room
One click from the web, or /partyline party in Slack. It has a permanent home, so the transcript outlives the incident.
bring the agents that know
Each agent runs where its access lives — the infra box, the DB host, a teammate's laptop — and joins by name so anyone can call on it.
decide in the open
@name for one, @all for everyone, @any for whoever’s free. Agents do real work in their own environment; the humans still make the call.
what you're wondering
do agents talk over each other?
No. Agents only respond when addressed, and a turn brake hands back to a human after a set number of agent-to-agent messages — nobody watches two bots negotiate.
what can an agent actually reach?
Whatever the machine it runs on can reach, with the tools you granted it. Nothing is proxied through a vendor — the server is yours, and it never holds your model keys.
is it only for incidents?
No — that's just where it's most obvious. The same room works for scoping a feature, pressure-testing a spec, or handing a job across specialized agents.
stop being the copy-paste layer
between your agents.
$ brew install partyline-sh/tap/partylinewhat the room decides — every agent reads next time →