Skip to main content
Use this tool to execute concrete UI interactions after you identify targets with Get Active App Context.
For multi-step UI work, prefer the Computer Use agent (computer_use_task), which calls perform and perform_sequence automatically. See the agent tools reference for batching rules.

​
Supported actions

  • click: Click a target element (default).
  • fill: Enter text into a text field or textarea identified by elementID.
  • type: Type raw keystrokes into whatever currently has keyboard focus — no elementID needed (useful for apps without a fillable field, e.g. typing a full expression into Calculator).
  • shortcut: Send keyboard shortcuts (for example cmd+c, ctrl+shift+t).

​
Inputs

  • action (optional): click, fill, type, or shortcut (default click).
  • elementID (required for click and fill): Element ID from Get Active App Context output.
  • text (required for fill and type): Value to insert or keystrokes to type.
  • shortcut (required for shortcut): Shortcut string.
  • windowID (optional): Target a specific window (pid/title) instead of the active one.

​
Notes

  • Element IDs expire with UI changes and session timing. Re-read context if an element is no longer valid.
  • For background window targeting, use the same windowID across context read and perform steps.

​
Good pairing