DGX Spark - NemoClaw agents restricted, cron jobs unscheduleable

[csanad.poda]

I freshly installed NemoClaw v0.0.55 as recommended by the official guide, after destroying previous sandboxes, but agents seem clobbered. I previously was able to just tell the agent to do something like execute a command or run a cron job, and it just did it. Now I keep facing “Tool exec not found” and the like, and especially, Cron jobs completely fail with an 1006 Gateway error:

{
“status”: “error”,
“tool”: “tool_search_code”,
“error”: “Error: gateway closed (1006 abnormal closure (no close frame)): no close reason\nGateway target: ws://127.0.0.1:18789\nSource: env OPENCLAW_GATEWAY_URL\nConfig: /sandbox/.openclaw/openclaw.json\n\nPossible causes:\n- Gateway not yet ready to accept connections (retry after a moment)\n- TLS mismatch (connecting with ws:// to a wss:// gateway, or vice versa)\n- Gateway crashed or was terminated unexpectedly\nRun `openclaw doctor` for diagnostics.”
}

I’m baffled because what was an okay system is now completely unusable. What am I missing here?

[elsaco]

The latest OS from 2026.05.31, release version 1.135.34, has this interesting note:

This month’s DGX Spark OS release brings the most streamlined out-of-the-box experience with a streamlined NemoClaw installer, along with faster inference on the top agentic models.

You might want to try this new image.

[csanad.poda]

“NemoClaw Playbook - After initial system boot, the DGX Spark playbook site opens, where the NemoClaw playbook is prominently displayed for streamlined setup of a sandboxed local agent”

That’s the playbook I’m using tho.