Claude Code plugin for Home Assistant - AI-assisted automation management

Share your Projects!
, , ,
1

Hey everyone,

Wanted to share something I’ve been working on that’s genuinely changed how I interact with my Home Assistant setup. Not a polished product announcement - more of a “here’s an interesting direction” post.

The short version:

I gave Claude Code (Anthropic’s AI CLI tool) direct access to my Home Assistant - SSH for commands, SSHFS mount for config files, and a custom MCP server that indexes entity dependencies. It can now read,
understand, and modify my entire setup with full context.

Why this feels significant:

I just overhauled ~80 automations that had accumulated over years. The AI could trace what used what, identify dead entities, spot near-duplicate automations I’d forgotten about, and validate changes against
my live instance before reloading.

Work that would’ve taken me a weekend of careful grepping and testing took an afternoon. Not because it wrote code faster - because it understood the whole system while making changes.

The bigger picture:

I can genuinely imagine a future where someone describes their home (“3 bedrooms, Hue lights, Ecobee thermostat, I want the lights to follow me and the heat to adjust when I leave”) and a system like this
generates the entire config - entities, automations, helpers, dashboard.

We’re absolutely not there yet. This is firmly power-user and coder territory right now. Giving an AI direct SSH access and file mounts means there’s serious potential for fuckups. You need to understand what
it’s doing, review changes, and know how to recover when things go wrong.

But the foundation is interesting:

  • Direct file access means it works with your config, not generic examples
  • Entity dependency tracking means it knows what breaks before changing things
  • SSH access means it can validate (ha core check) and reload without manual steps
  • MCP protocol means you can extend it with your own tools

What I built:

ha-pilot - a Claude Code plugin with:

  • Skills containing HA-specific YAML best practices, API reference, diagnostics guides
  • An MCP server (Python) that indexes your YAML config and builds a dependency graph
  • Setup wizard for SSH, SSHFS, environment configuration

Links:

This isn’t ready for “install and forget” users. But if you’re comfortable with the terminal, curious about where AI-assisted home automation might go, and willing to experiment - I’d love to hear your
thoughts.

Anyone else exploring this space?

2 Likes
2

Yes! I’ve been doing something similar, but mainly (so far) focused on a different host where I have a lot of different services running, including some that are related to home assistant… and that’s been extremely effective, so much so that I added a couple of SMB mounts to access Home Assistant’s config, and a couple of MCP servers (plus an API token) to be able to poke the server directly – and for a few tasks so far it’s been just as effective & productive as you’re describing.

I can tell I need to get better & more up-to-date documentatiron into the context window via skills or just having them on-hand – some of what i want to do is modernize e.g. some old light group entities to instead (or in addition) be built using labels - monotonous stuff, but that requires more than a gut-feel for what a homeassistant yaml file wants to look like vs a few years ago.

Nonetheless – feels very very promising!