Coordinate work that takes weeks

What happens

1. Divide the responsibility

   dev-lead@factory assigns the interface to dev-ui@factory and the backend to dev-app@factory. Each assignment gets its own tracked task. The specialists read and claim their tasks, so there is a named owner for each part.

2. Use temporary helpers inside each assignment

   Each specialist delegates two small investigations to its own subagents. These helpers appear inside the parent’s terminal. They return findings about existing code and tests; they do not become new members of the rig.

3. Connect the two pieces

   dev-ui@factory asks dev-app@factory which backend connection to use. The answer incorporates the backend investigation: reuse the shared client and its existing reconnect support. The specialists coordinate directly instead of sending every detail through the lead.

4. Keep the larger effort moving

   The scene ends with both assignments underway. dev-lead@factory will coordinate progress and bring the results together for review. The weeks of implementation are ahead; this exchange shows how the team starts them.

What OpenRig does

OpenRig supplies named seats, tracked assignments through rig queue, and direct communication through rig send. The interface and backend agents remain separate teammates with their own terminals and responsibilities.

What Claude or Codex does

Codex and Claude Code perform the reasoning, use tools and delegate the bounded investigations to their native subagents. Those temporary helpers belong to a parent’s current job. OpenRig’s named teammates can keep receiving work and coordinating beyond that one delegation.

When this helps

A large project accumulates more decisions and unfinished work than one conversation can comfortably hold. Dividing responsibility lets specialists concentrate on their areas while a lead brings their work back to the developer.