OnboardIQ
Watch the agent read a device manual, test its picks on the device, and correct itself.
These are replays of recorded live sessions. Every reading and every model answer comes from the recording; the verifier recomputes each check as you watch.
Start here
- 1. Watch it correct itself Onboards a device from its manual, tests each pick on the device, fixes a wrong one.
- 2. Watch a health score react CPU stress on a real machine: the score falls 100 → 40, then recovers, with a plain-English explanation.
- 3. Run it yourself Point it at your own device from a clone of the repo; setup steps are in the README. Open the README
Run it live
This runs the real agent now: Muse Spark reads the MIB text of a Linux host's net-snmp agent, every pick is tested on that device (inside this container, read-only), and CPU and memory are proven with a real load test. It takes about 3 minutes and costs real model calls, so it is limited: one run at a time, a pause between runs, and a daily cap. Nothing you type is sent to the device.
Run live
This page is connected to a local OnboardIQ server, so it can onboard a device now. The model reads the MIB modules you choose, then every pick is tested on the device.
Monitor live
Poll a device you have onboarded and score its health as it runs. Only the picks that were proven are polled; nothing is written to the device.
Choose a session
Pick a recorded session above to play it.