neeltom92/claude-code-observability — reverse-engineered prompt

Build me a simple local monitoring setup for Claude Code that I can start with one command on my Mac using Docker Desktop. I want it to spin up Grafana, Prometheus, Loki, and an OpenTelemetry collector, then automatically wire Claude Code into it by updating my Claude settings and making a backup first.

When it is running, I should be able to open Grafana on localhost and see ready made dashboards for cost, token usage, sessions, active time, model usage, code churn, tool activity, and skills usage. Keep everything bound to my machine only, not exposed to the network.

Please include basic checks so it confirms Docker is installed, the needed ports are free, the services are healthy, and it prints the local URLs. I also want easy commands to check status, view logs, restart, stop, and fully clean everything up, including removing the Claude settings patch. Please keep it self contained and local only, and look up current docs if you need to.