Member-only story
OpenRig Is Docker Compose for Coding Agents
Compose took containers that already existed and made a group of them describable and bootable as one system.
OpenRig does the same for coding agents.
Claude Code, Codex, and Pi stay exactly as they are, and OpenRig treats them as runtimes inside a topology you can define, launch, inspect, stop, and restore.
The project has converged on the following idea: a coding agent should occupy a seat in a system, with a role and an address that outlive any single terminal.
Take that idea seriously and it changes how you design agentic development workflows.
The Missing Layer Between Model and Team
The lower layers already have good abstractions.
A model gives you inference. Claude Code, Codex, and Pi wrap models with tool use, repository context, terminal access, and a coding loop.
As soon as you want two or five of those agents working together, though, most teams fall back to shell scripts, ad hoc prompts, and a human doing the coordination.
OpenRig adds the next layer up: the rig.
A rig defines which agents exist, which runtime each one uses, how they relate to each other, what role and context each one receives, and how…
