vintik100/bunbrowser

PRIMARY INSPECTION TOOL. Captures the complete semantic accessibility tree of the active page with deterministic element references ([ref=eN]). ALWAYS prefer this over screenshots for discovering interactive elements, reading page text, and finding buttons/inputs.

VISUAL ONLY. Captures an image screenshot of the current page viewport in Base64 encoding. Use ONLY for visual design validation, canvas/image inspection, or when explicitly requested. For element discovery and reading content, use browser_snapshot instead.

Execute an arbitrary JavaScript expression in the context of the active page and serialize its return value.

Retrieve raw HTML DOM source code or plain visible text of the active page.

Retrieve captured console logs (log, warn, error, info) emitted by the page JavaScript context.

Click on an element. PREFER using 'ref' (e.g. 'e1' from browser_snapshot) for reliable deterministic targeting. Fall back to CSS selector or coordinates (x, y) if ref is unavailable.

Type text into an editable input or textarea element. Prefer using 'ref' (e.g. 'e2'). For filling multiple form fields, use browser_fill_form instead to save steps.

BATCH FORM FILLER. PREFER over multiple individual browser_type calls. Fills multiple inputs and optionally clicks a submit button in a single turn.

Press a keyboard key or key combination (e.g. 'Enter', 'Escape', 'Tab', 'ArrowDown', 'Backspace', 'KeyA')

Hover the mouse pointer over an element to trigger hover states or dropdown menus

Scroll the viewport or scroll an element into view

Select one or more options in a <select> dropdown element

Drag an element and drop it onto another element

Extract real-time web performance metrics, Core Web Vitals (TTFB, FCP, load duration), JavaScript heap memory consumption, and network resource transfer breakdown.

FULL WEB QUALITY AUDIT. Computes Lighthouse scores (0-100) for Performance, Accessibility, Best Practices, and SEO. Returns actionable recommendations and diagnostics to optimize Core Web Vitals (FCP, LCP, TTFB, CLS) and user experience.

Navigate the active browser tab to a specified URL and wait for load

Navigate back in the browser history

Navigate forward in the browser history

Reload the current page in the active tab

Execute a raw Chrome DevTools Protocol (CDP) command on the active tab (Chrome/Chromium backend)

Get, set, or clear cookies for the active browser session

Inspect, set, or clear localStorage keys in the active page origin

List all open browser tabs with their IDs, URLs, titles, and active status

Open a new browser tab/view, optionally navigating to an initial URL

Switch the active browser context to a different tab by tabId

Close a browser tab (closes the active tab if tabId is omitted)

Resize the viewport dimensions of the active browser tab

CONTINUOUS VIDEO RECORDER. Starts background video recording of user interactions and animations on the active tab into WebM or animated GIF. Use 'browser_stop_recording' when done to export.

Stops the active video recording, exports the resulting WebM or animated GIF file to disk, and returns duration and frame metrics.

ONE-SHOT ANIMATION & MOTION RECORDER. Records UI transitions, CSS keyframe animations, micro-interactions, or visual jank for an exact duration (durationMs) into WebM or animated GIF. Supports triggering actions via triggerScript or triggerSelector.