The three layers
TapKit has three pieces that work together: your iPhone, the Mac app, and the TapKit server.The Mac app is the host
The Mac app runs on a Mac that’s physically connected to your iPhone via USB. It receives the phone’s screen and sends taps, swipes, text, and device commands through the paired phone connection. The Mac is the only TapKit component that talks directly to the phone.The server connects everything
The Mac app maintains a persistent connection to the TapKit server. When an AI agent sends a command through the API — “tap at (200, 400)” or “take a screenshot” — the server routes it to the right Mac, which executes it on the right phone. The server also handles authentication, session management, and screen streaming.Your agent calls the API
Any tool that can make HTTP requests or connect to an MCP server can control the phone. Claude, Codex, your own scripts — they all talk to the TapKit API, never directly to the Mac or the phone. This means your agent doesn’t need to be on the same network, or even the same continent, as the device it’s controlling.Why this architecture?
- No jailbreak, no sideloading, no developer mode. Nothing is installed on the iPhone.
- Works with any app. TapKit interacts with the phone through the same visible interface a person uses, rather than relying on app-specific integrations.
- Remote by default. The API layer means your agent can be anywhere. The Mac and phone can be in a closet, a data center, or on your desk.
What happens when you run a task
- Your agent sends a command to the TapKit API (or MCP server).
- The server authenticates the request and routes it to the Mac that’s hosting the target phone.
- The Mac app executes the action on the phone.
- The Mac captures the updated screen and sends it back through the server.
- Your agent sees the result and decides what to do next.