Create an isolated browser session. Each session has its own cookies, storage, and tabs; sessions never share state, so many agents can drive separate sessions at once.

List the ids of all live browser sessions.

Close a session and release its browser context.

Save the session's cookies + localStorage to a JSON file (within the output dir) for later reuse via browser_create_session.

Navigate the session's active page to a URL.

Go back to the previous page in the session.

Click an element (by ref from browser_snapshot).

Fill an input with text (by ref from browser_snapshot).

Hover over an element (by ref from browser_snapshot).

Press a keyboard key (e.g. "Enter", "ArrowDown").

Select one or more options in a (by ref from browser_snapshot).

Fill several fields in one call (each by ref from browser_snapshot).

Set files on a file input (by ref from browser_snapshot).

Drag one element onto another (both by ref from browser_snapshot).

Resize the session's viewport.

Capture a PNG of the active page and return it as an image. Optionally also save it to a file within the output dir.

Get the accessibility snapshot (YAML) of the active page, including element refs to target with other tools.

Run a JavaScript expression in the page and return the JSON-serialized result.

Wait until a CSS selector reaches a state.

Return console messages captured in the session.

Return network responses captured in the session.

Decide the next dialog (alert/confirm/prompt) instead of auto-dismissing it.

List, open, close, or select a tab within the session.