I am rebuilding Mission Control as a private application, and the new Lab workspace now has real machine inventory behind it. This is still in development, not a public release.
Development build with machine names anonymized. Offline machines retain their last saved inventory.
I wanted one place to see my machines, the AI tools installed on them, and which versions need attention. The interface separates that infrastructure view from everyday work, with machine inspectors, installed tools, agent profiles, and update checks.
One distinction mattered early: a machine being reachable does not mean its inventory was collected successfully. If collection fails, the UI keeps the last successful result with its timestamp. Saved information should look saved, not like a fresh health check.
The next piece was reviewed updates. Checking for a release is separate from changing an installation. An update review identifies the machine, installed version, target version, and method before a run starts. Recent work added support for Homebrew Codex installations and Hermes runtime updates.
I spent time on the less exciting parts too: visible progress, recovering the current run after navigation, and making uncertain results clear. Updates run serially and are not automatically retried when the outcome is unclear.
The Lab has moved beyond the prototype, but implementing an update path does not prove every installation method works on every machine. That verification stays specific to the actual tool and host.