•23d ago
What would stop you from self-hosting your RMM?
ET Ducky Local-First runs the whole RMM on your own infrastructure with Docker Compose, with unlimited seats and technicians at one flat price. It isn't completely air-gapped. Sign-in goes through Clerk, and the license renews against etducky.com.
If you run an RMM on your own hardware, which outbound connection would fail your security review? Is the sign-in path or the license check the harder one to get approved?
Drafted with AI, reviewed by me.
•23d ago
Hypership: what should I build on launch day?
ET Ducky is in Product Hunt's Hypership challenge on Wednesday, September 30. I'll read every feature request in the launch comments and ship what I can that day.
The agent reads the kernel event stream on Windows (ETW) and Linux (eBPF), and every alert gets a root-cause analysis built from those events. Remote desktop, patch management, behavioral security, and reports run from the same agent.
What's missing from the RMM you use today that a kernel event stream could answer? Post it here or in the launch comments tomorrow.
Drafted with AI, reviewed by me.
•23d ago
How do you decide a patch ring is safe to promote?
ET Ducky marks a patch as successful only when the kernel event stream shows services back up, no crash loops, no deviation from that machine's own baseline, and no behavioral detections. Rollout rings use tags and a soak time in hours, and a regressed result pauses the deployment before the next ring.
How long do you let a first ring soak before you promote it? Which signal would make you pause a rollout that your patch report already shows as installed?
ET Ducky launches here on Wednesday, September 30. I'll read every reply.
Drafted with AI, reviewed by me.
•23d ago
Would you let an AI choose what to trace on a production server?
In an ET Ducky Live Session the AI picks which ETW providers to turn on for each round of an investigation. Only counts-only summaries leave the host, and the raw kernel events stay on the endpoint. You can use your own Anthropic or OpenAI key on any tier.
Guided Troubleshooting works the other way. The AI proposes one diagnostic action at a time, and nothing runs until a person approves it.
What would you need to see before you let an AI choose what to trace on a production server? Is a log of the providers it picked each round enough, or would you want to approve every round?
Drafted with AI, reviewed by me.
•23d ago