OpenRig explained: coordinate coding agents without losing the operator
Deploying OpenRig: separate instance state from provider state
Plan a controlled workstation or shared-host rollout before starting seats
What you will learn
- Locate state and resources
- Pin the runtime and launch path
- Make stopping and recovery explicit
Before you start
- Node 22 or 24 and tmux on a supported host
- One working provider account and a disposable repository
Turn the first owner/checker run into evidence for a safe rollout decision
Key takeaways
- OPENRIG_HOME does not isolate provider settings.
- Repository docs may be ahead of an npm release.
- Recovery must account for per-seat outcomes.
Locate state and resources
The daemon stores its database and instance resources under `OPENRIG_HOME`, normally `~/.openrig`. Managed sessions also write provider-specific configuration and workspace-local files outside that directory.
Back up the exact tmux, Claude and Codex paths listed in the README. Pointing `OPENRIG_HOME` at a new folder is not a complete isolation plan because provider hooks and trust entries can still change.
Pin the runtime and launch path
Choose Node 22 or 24, the CLI release and a known tmux version. The project warns that repository guidance may be ahead of npm, so compare `rig --version` with the documentation you use.
For a team host, decide who owns the daemon account, provider credentials and workspace. Avoid treating a shared shell account as an access-control boundary. Validate the selected provider login before a rig starts.
Make stopping and recovery explicit
A `rig down --snapshot` operation records topology for later restoration; `rig up <name>` can restore from a snapshot or current database state. The restore report distinguishes resumed, fresh and failed nodes.
Test the stop and restore path on a disposable rig. An upgrade has its own procedure; the README explicitly says not to use `rig down` as a casual upgrade step for live seats.
Decision guide
| Criterion | Option A | Option B |
|---|---|---|
| Best when | You need predictable behavior and easy auditing | You need adaptive optimization and have reliable telemetry |
| Main risk | May leave performance on the table | Can become difficult to explain or debug |
Implementation steps
- 1
Inventory instance, provider and workspace write paths.
- 2
Pin Node, CLI and provider versions for the pilot.
- 3
Rehearse stop, snapshot and restore with disposable seats.
Copy-ready example
pilot:
node: 22-or-24
cli_release: pinned
provider: selected-and-authenticated
config_backup: verified
restore_trial: disposable-rigFrequently asked questions
Is moving OPENRIG_HOME enough for a clean test?
No. Provider trust, hooks and workspace resources can be written elsewhere.
Should I stop every seat before upgrading?
The pinned upgrade guidance says `rig down` is not the upgrade step.
Sources
- OpenRig / README.mdSource checked 2026-10-04
- OpenRig / docs/reference/instance-layout.mdSource checked 2026-10-04
- OpenRig / docs/reference/host-resources.mdSource checked 2026-10-04
- OpenRig / docs/reference/getting-started.mdSource checked 2026-10-04
- OpenRig / packages/cli/src/commands/down.tsSource checked 2026-10-04