OpenRig vs OpenClaw

Why I publish comparisons. This category is new. The vocabulary isn't settled. Nobody has a complete mental model of what's in the space, including me. The comparisons here are an attempt to find the shape of the category — what approaches are being taken, what each approach looks like, what we might learn from each other. Not a bake-off.

OpenClaw started from: what if an agent had a stable identity — a personality, a role, a set of values — and we built a whole methodology around composing those entities? The answer is an agent-first system. Each OpenClaw agent has a SOUL.md file that defines who it is. The community has built hundreds of these agents, each shaped for a specific role. When OpenClaw needs to reach outside itself, ACP (Agent Client Protocol) lets it spawn Claude Code, Codex, or other coding harnesses as tools. The unit of work is the agent.

OpenRig started from a different question: what if Claude Code and Codex are already extraordinarily capable, and what I actually want is to run several of them together as a coordinated team? The answer is a topology-first system. OpenRig doesn't build its own agents and doesn't give them identities — it runs Claude Code and Codex as they ship, unchanged, arranged into a rig you define in YAML. The unit of work is the rig.

Neither answer is obviously right. They're different bets about where the leverage is in multi-agent coding.

Because a rig is just a set of real Claude Code and Codex sessions with defined relationships, a few things fall out naturally:

• You boot 8 agents with one command. rig up demo gives you orchestration, development, and review pods, all running Claude Code and Codex, ready to take work. You don't design a topology from scratch — the spec library ships proven shapes you can start from.
• You save the whole rig and come back to it. rig down snapshots the topology; rig up <name> brings it back. Your agents were in the middle of something — they're back in the middle of it.
• You bring in agents you're already running. rig discover finds Claude Code and Codex sessions you launched manually in tmux; rig adopt binds them into a rig. Your existing work becomes managed, without being rebuilt.
• You extend a rig with services. A rig can ship actual software alongside a specialist agent whose job is to run it. The shipped example is HashiCorp Vault with a vault-specialist agent — one command and you have a secrets manager living inside your rig. You can compose your own services the same way.

Because agents are Claude Code and Codex as they ship, you use them the way you already do. You attach to a pane, you talk to Claude or Codex in its TUI, it runs commands, writes code, the usual. The difference is that now it has peers — other agents it can send messages to, ask for help, or hand work off to.