Docs / Going further

Agents that work together

Every agent knows it's running inside Flotilla. Ask it to hand off a side task, start a helper or message another agent, and you'll see it all happen in the sidebar.

Just ask

You don't need to learn anything new. Ask an agent in plain words:

"Hand off the flaky test fix to a new agent and keep going on the feature."
"Start an agent to write tests for this once you're done."
"Tell the api-auth agent that the endpoint name changed."
"Let me know when the build is green." The agent sends you a notification.
"Set up a routine that runs the full test suite every night."

New agents appear in the sidebar like any other, so you stay in charge. You can open, redirect or stop any of them.

How agents do it

Agents use a command called flotilla that comes with the app. You can use it too, from any agent's Terminal tab:

`flotilla agents`	List every agent and what it's doing.
`flotilla new fix-login --prompt "…"`	Start a new agent on a task.
`flotilla fork`	Start a copy of this agent, conversation included.
`flotilla send api-auth "…"`	Message another agent.
`flotilla notify "…"`	Send yourself a notification.
`flotilla routine add …`	Schedule a routine.

Run flotilla on its own to see everything it can do.

What agents are told

Each Claude Code and Codex agent starts with a few house rules from Flotilla: stay in its own checkout, keep one pull request going at a time, open pull requests as drafts, and lead its replies with the outcome. That's what keeps a fleet of agents tidy. You can add your own instructions in Settings › Prompt; see Fewer interruptions.