MCP

TapKit runs a hosted MCP server. Any client that supports remote MCP servers with OAuth can connect to it. The only thing you run locally is the Mac app.

https://mcp.tapkit.ai/mcp

The server uses the Streamable HTTP transport.

Connect

Most clients accept a config like this:

{
  "mcpServers": {
    "tapkit": {
      "type": "http",
      "url": "https://mcp.tapkit.ai/mcp"
    }
  }
}

Point your MCP client at the URL. The client must support OAuth 2.1 with PKCE and dynamic client registration. The server publishes its OAuth metadata at /.well-known/oauth-protected-resource/mcp. The first time your client connects, it opens TapKit in your browser. Sign in with Google, then click Allow access to let the client view and interact with your connected iPhones. The MCP server only accepts OAuth. API keys are for the REST API.

Tools

Start with list_phones. Every other tool takes the phone_id it returns.

Tool	Parameters	Description
`list_phones`	None	List phones with their IDs, status, and screen size
`get_phone_status`	`phone_id`	Connection status and screen size of one phone
`screenshot`	`phone_id`	Capture the screen as an image
`tap`	`phone_id`, `x`, `y`	Tap a point
`double_tap`	`phone_id`, `x`, `y`	Double tap, for zooming or selecting a word
`triple_tap`	`phone_id`, `x`, `y`	Triple tap, for selecting a line or paragraph
`long_press`	`phone_id`, `x`, `y`, `duration`?	Press and hold, for context menus. `duration` is in ms, default 1000
`flick`	`phone_id`, `x`, `y`, `direction`	Quick swipe `up`, `down`, `left`, or `right` from a point
`drag`	`phone_id`, `from_x`, `from_y`, `to_x`, `to_y`, `speed`?	Press, move, and release. `speed` is `slow`, `medium`, or `fast`, default `medium`
`hold_and_drag`	`phone_id`, `from_x`, `from_y`, `to_x`, `to_y`, `hold_duration_ms`?, `speed`?	Long press, then drag, for moving items. Hold defaults to 500 ms
`type_text`	`phone_id`, `text`	Type into the focused field. US-ASCII only, up to 100 characters
`press_key`	`phone_id`, `key`, `modifiers`?, `repeat`?	Press `enter`, `escape`, `backspace`, or an arrow key, optionally holding `control`, `shift`, `alternate`, or `command`
`press_home`	`phone_id`	Go to the Home Screen

Coordinates

Screenshots are JPEGs scaled so the long edge is at most 1344 pixels. Tool coordinates are pixels in that image, measured from the top-left corner. The server converts them to the phone’s native resolution. Action tools don’t return a screenshot. Call screenshot after an action to see the result.

Limits

Each connection can make 300 requests per minute. Agents can only control phones that are online and active on your subscription.

Getting started

Anthropic

OpenAI

Everywhere else

Connect

Tools

Coordinates

Limits

​Connect

​Sign in

​Tools

​Coordinates

​Limits

Connect

Sign in

Tools

Coordinates

Limits