Skip to main content
Any tool that speaks MCP (Model Context Protocol) can connect to TapKit. This includes Cursor, Windsurf, custom agents, and any other MCP-compatible client.

​
Setup

Add the TapKit MCP server to your client’s configuration: 
{
  "mcpServers": {
    "tapkit": {
      "type": "http",
      "url": "https://mcp.tapkit.ai/mcp"
    }
  }
}
Where you add this depends on your client:
  • Cursor — add to your MCP settings in the Cursor configuration
  • Windsurf — add to the MCP configuration file
  • Custom clients — point your MCP client at https://mcp.tapkit.ai/mcp

​
Authentication

When your client first connects to the TapKit MCP server, you’ll be prompted to authenticate via browser. Sign in with your TapKit account and the token will be cached for future sessions.

​
Available tools

Once connected, your client has access to all TapKit tools:
ToolDescription
tap(x, y)Single tap at coordinates
double_tap(x, y)Double tap
long_press(x, y)Tap and hold
swipe(x, y, direction)Flick gesture
drag(from_x, from_y, to_x, to_y)Drag between points
hold_and_drag(from_x, from_y, to_x, to_y)Long press then drag
screenshotCapture current screen
open_app(app_name)Open any app by name
type_text(text)Type into focused field
press_homeGo to home screen
escapeDismiss keyboards/alerts
lock / unlockLock or unlock device
volume_up / volume_downVolume controls
spotlight(query?)Open Spotlight search
activate_siriTrigger Siri
get_phone_infoGet screen dimensions
list_phonesList connected phones
select_phone(phone_id)Select a specific phone

​
How it works

The MCP server runs as a hosted endpoint at mcp.tapkit.ai. You don’t need to run anything locally beyond the TapKit Mac app with a connected iPhone. Your MCP client connects to the server, which routes commands to the Mac app, which executes them on the phone. Screenshots are returned through the same channel. All coordinates map 1:1 with screenshot pixels. Screenshots are scaled so the longest edge is 1344px — the server handles coordinate conversion transparently.

​
Tips

  • Auto phone selection — if only one phone is connected, it’s selected automatically. No need to call list_phones first.
  • Multi-phone setups — every tool accepts an optional phone_id parameter for targeting a specific phone.
  • Text input requires focus — tap a text field before calling type_text.
  • The agent loop — the standard pattern is: screenshot → reason about what’s on screen → perform action → screenshot again to verify.