Tell your AI assistant you are heading to bed and have it turn off lights, dim lamps, adjust the thermostat, and lock the front door.
Use the command line tool to list and control all your Apple Home accessories from the terminal without opening the Home app.
Build an AI agent that can activate or create scenes and run automations across your Apple Home setup.
Expose your Apple Home devices to AI clients like Claude or Cursor through a structured tool server.
| liiraliu/homekit | diendh/zca-bridge | utibeabasi6/mercek | |
|---|---|---|---|
| Stars | 69 | 69 | 69 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 3/5 | 4/5 | 3/5 |
| Audience | developer | developer | ops devops |
Figures from each repo's GitHub metadata at analysis time.
Requires macOS 13 Ventura or newer, a configured Apple Home on the same Mac, and authorization through the Homekit macOS app.
Homekit is an open source project that lets AI assistants control Apple Home devices. Instead of tapping buttons in the Home app or asking Siri, you can have an AI agent turn lights on and off, lock doors, read sensor data, and trigger scenes or automations. The project positions itself as the first open bridge between Apple's HomeKit framework and the AI agent layer. The project is built in TypeScript and ships as several npm packages. The command line tool, homekit-cli, gives you a terminal interface to list accessories, set their states, activate scenes, and run automations. There is also homekit-mcp, a server that exposes eight structured tools any compatible AI client can call. A shared core package handles authorization, types, and the bridge logic, and an optional plugin supports a framework called OpenClaw. The architecture has three layers. At the top, an AI agent like Claude or Cursor talks to the homekit-mcp server using a protocol called MCP. That server communicates over a local socket with a macOS app that handles authorization and talks to Apple's HomeKit framework. From there, it reaches your actual devices: lights, locks, thermostats, sensors, cameras, scenes, automations, and multi-home setups. The eight tools an agent can call cover listing accessories, getting and setting accessory state, activating and creating scenes, and listing and running automations. The README includes a demo conversation where someone tells Claude they are heading to bed, and the agent checks what is on, turns off the right lights, dims the bedroom lamp, adjusts the thermostat, and locks the front door. Requirements are macOS 13 Ventura or newer, Node.js 18 or newer, a configured Apple Home on the same Mac, and the Homekit macOS app for initial authorization. Installation is a global npm install followed by an auth step. The full README is longer than what was shown.
Lets AI assistants like Claude control Apple Home devices, lights, locks, thermostats, through a local bridge so your agent can manage your smart home by conversation.
Mainly TypeScript. The stack also includes TypeScript, Node.js, npm.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.