browser_take_snapshot | The page as one line per element, with a number on every element the agent can act on, plus any popups in the way. Can add an annotated screenshot or read beyond the screen. |
browser_read_pages | Reads up to 10 pages at once in background tabs and returns their text, links included. |
browser_batch | Several clicks, typing and key presses on one tab in one call, then a fresh snapshot. |
browser_click, browser_input_text, browser_press_key, browser_select_editable_text, browser_scroll | Act on an element by its number. |
browser_move, browser_drag | Hover, and drag sliders, canvases and reorderable lists. |
browser_navigate, browser_go_back, browser_go_forward, browser_reload, browser_stop_loading, browser_wait | Getting around. |
browser_new_tab, browser_select_tab, browser_close_tab, browser_set_audio_muted | Tabs. |
browser_get_window_info, browser_new_window, browser_close_window, browser_show, browser_minimize | Browser windows. |
browser_close_popup_bubble | Answers a permission prompt or dialog. |
browser_start_human_control, browser_stop_human_control | Hands the tab to you for a password or captcha, then takes it back. |
browser_status | Whether the browser is ready, or how far its first install has got. |
browser_run_task, browser_stop_task | Optional: hands a whole task to a vision model at an endpoint you set up, with your own API key, and keeps the screenshots out of your agent’s context. |