Yes, Dave. I can do that
I like keeping things simple and easy since I'm only one person but a solo developer, new dad, I have client work and Half Systems products on the side to manage. I take help where I can get it.
Over the last year LLMs have been getting very good (such an insight, I know) at aiding in debugging. If you need a quick SQL query or a fast analysis of a large stack trace from a crummy PHP app, you can get answers quick.
The start of any of these interactions though is get the logs. My services run in multiple containers on a remote box. So it's SSH in, podman logs the right pod, copy through tmux (which can be a pain with panes), paste into ChatGPT, wait for the tab to recover from whatever it's doing, and then ask your question. Depending on the urgency of the error, it's a bit much and error prone for "why did this request fail."
Last year as I started to build out the new services for Half Systems, I learnt about Model Context Protocol. It seemed interesting but it wasn't until the start of this year that it clicked. Adding an MCP endpoint to my services allows the model to investigate internal state and logs of a service. No copying. No mangled clipboard. It just asks the service directly.
Now I can just ask and it does the leg work.
❯ Can you check the status of our services
⏺ ifnot200 - service_health (MCP)
⎿ [
{
"name": "Database",
… +15 lines (ctrl+o to expand)
⏺ ifnot200 - system_stats (MCP)
⎿ {
"total_targets": 8,
"checks_today": 7242,
… +2 lines (ctrl+o to expand)
⏺ ifnot200 - pod_logs (MCP)(container: "api", since: "1h", tail: 50)
⎿ ifnot200-api: 2026/03/05 10:17:55 scheduler.go:117: checking 1 targets
ifnot200-api: 2026/03/05 10:17:56 scheduler.go:117: checking 1 targets
ifnot200-api: 2026/03/05 10:17:59 scheduler.go:117: checking 2 targets
… +28 lines (ctrl+o to expand)
⏺ All services are healthy
As always, I appreciate any feedback or if you want to reach out, I'm
@xlvdk or @neuralsandwich on twitter and most other places.