I wrote an open-source MCP server that you can run locally and connect to an AI client like Claude Desktop.
The repo is here: https://github.com/juhapellotsalo/ruuvi-mcp-server.
The idea is to run the server on your local network (e.g., on a laptop, Raspberry Pi, etc.). It first needs to pull the sensor readings into an SQLite database.
I built several mechanisms for that: syncing from the cloud account or getting the 10-day history directly from the device over BT. Then you can opt to listen to BLE or MQTT broadcasts or just poll the Gateway over HTTP. It stores all readings in an SQLite file.
You can then configure a local MCP client to talk to the server. I used Claude Desktop since it’s the easiest to set up, and then you can ask things like “analyze my office air quality from the last week.” Claude can draw all sorts of graph artifacts that are quite interesting to play with.
There’s also a config file (devices.yaml) where you can give a description to the devices, like “This Air is in my office. No mechanical ventilation, just open the windows sometimes.” This information gets passed to the AI client, giving it more context what the raw sensor readings mean.