Tested my first MCP server: kernelgarage.mcp_server, built to report hardware + LLM usage from the same Prometheus stack...

Tested my first MCP server: kernelgarage.mcp_server, built to report hardware + LLM usage from the same Prometheus stack as the last two episodes. Hooked it into Claude Code, asked in plain English, worked first try. A follow-up for the last 7 days made it notice something: peak queue depth of 19, average wait climbing to ~18 minutes. It didn't just answer, it flagged a real problem, unprompted. Any experiences with using MCP servers in your homelab?#Python #LLM #OpenSource #Homelab #MCP

Read Original

Related