Runtime Communications

Coordinate multi-agent ML research workflows using tmux by reading shared runtime threads, inspecting per-agent inboxes, and delegating tasks across leader, researcher, and trainer agents via the runtime CLI.

An agent inbox in a fixed-agent runtime is a per-agent message queue that you inspect to assign tasks and track progress. It enables explicit task delegation and handoffs between leader, researcher, and trainer roles.

Delegate tasks across leader, researcher, and trainer agents using the runtime CLI and direct messages. You read shared threads to understand project state, then inspect per-agent inboxes to assign and monitor delegated work.

Yes, the runtime communication contract supports human-in-the-loop workflows in local tmux-based environments. It enforces governance rules described in AGENTS.md and runtime.toml to maintain consistency during multi-agent task delegation.

Yes, you need AGENTS.md and runtime.toml configuration files to enforce multi-agent runtime governance rules. These files define the communication contract and ensure consistent task handoffs and visibility across the fixed-agent workflow.