OpenRod Cloud is live: one-click @NVIDIAAI OpenShell sandboxes, for FREE.
2,600 downloads in our first 3 days, and some of the feedback (thanks, Reddit) asked for cloud hosted OpenShells.
So we built it.
Sandboxes are deny-by-design, so the hard part is writing the right policy.
Agents keep changing, and writing the right policy for each one at scale is close to impossible.
What's the hardest part of writing policy for your agents?
`npx openrod`
OpenRod is live.
OpenRod is an open source control plane for @nvidia OpenShell.
Spin up a sandbox for any agent you want.
Run it locally or over SSH on a host you control.
Set egress, attach MCPs and skills, and keep an audit trail of what it did.
`npx openrod`