Use a background agent with inbox-style approvals

They used a desktop app to run agents in the background and review results later in separate inboxes for finished work and decisions.

Done with Custom agent· Intermediate

What they did
They launched tasks in a self-hosted desktop app that also works in a browser or through the command line. The agent kept going after they moved on, with each run saved as a checkpoint. When a run finished, it landed in Unread; if it needed a choice, it showed up in Action. They could also start runs on a schedule or from a webhook, and give agents access to local folders only when allowed.
How it went
The setup was designed for background work, but the only usage figure mentioned is that more than 2,000 Amazon users tried earlier versions. The available material does not confirm the CRM logging and meeting prep example.
Tips
Keep the server process running if you want long-running jobs to continue. Make sure Node.js 24+ is installed and a model provider is set in Settings > Providers before you start.

Try it yourself with OpenClaw

Each [weekday morning], use my connected [calendar] and [CRM] to draft [meeting prep notes] for upcoming meetings and [activity logs] for yesterday’s calls. Put completed drafts in an unread inbox, and put anything that would update the CRM, send a message, delete anything, or submit anything in an action inbox. Pause and ask for my approval before doing any of those irreversible steps.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.