Player not loading? Watch on YouTube
This overview presents the speaker's preferred MCP servers for home lab administration and infrastructure as code in 2026. It describes how an AI agent can use MCP to interact with applications and infrastructure, with examples of conversational queries rather than a detailed installation walkthrough.
Proxmox MCP connects through the official API. The speaker describes checking running virtual machines, available node memory and stopped LXC containers, plus creating a VM from a template. GitHub MCP supports questions about commits, branches and repository workflows; the speaker says it runs in Docker. Filesystem MCP helps explain projects spread across directories and YAML files, with access limited to directories the user explicitly permits.
The Kubernetes examples cover restarting pods, unhealthy deployments, namespace events and CPU usage. The speaker says he still verifies important production changes. Context7 supplies current documentation and code examples for a coding assistant instead of relying solely on model training data.
The final section discusses jcodemunch, jdocmunch and jdatamunch, focusing on jcodemunch's project index. The speaker reports faster responses and reduced cloud token use, but advises testing the savings independently because the tools report their own estimates.
