Browser tool

Copy currently selected text on the page to clipboard - ideal for extracting highlighted content, copying form data, or harvesting visible text selections.

Execute precise drag and drop operations - essential for file uploads, list reordering, element moving, and complex UI interactions that require drag-based manipulation.

Read page content - essential before clicking. Use BEFORE clicking to: - Find CSS selectors for elements - Understand page layout - Verify current state FORMAT OPTIONS: - 'html': Raw HTML - use to find selectors for MouseClick - 'markdown': Clean text - use for reading/analysis WORKFLOW: FetchWebpage(format='html') -> find selector -> MouseClick(selector) WHEN TO USE: - Before any click to find the right selector - After actions to verify they worked - When you need to read page content - As fallback when screenshot fails

Read current content from the system clipboard - essential for data transfer workflows, extracting copied text, and reading user-copied data for processing.

Execute keyboard shortcuts and key combinations - essential for copy/paste, navigation, and application commands that agents need for efficient browser automation.

Click element by CSS selector. WORKFLOW: 1. FetchWebpage(format='html') to see page elements 2. Find selector for target element 3. MouseClick(selector='#your-element') SELECTOR EXAMPLES: - By ID: '#submit-button' - By class: '.btn-primary' - By attribute: 'button[type="submit"]' - By text: 'a:has-text("Login")' - Combined: 'form#login button.submit' IF CLICK FAILS: 1. FetchWebpage to verify element exists 2. Check if element is visible 3. Try more specific selector 4. Use PerformWebTask for complex interactions

Execute a precise double click at specified screen coordinates - ideal for opening files, selecting text, or activating UI elements that require double click gestures.

Press and hold mouse button at coordinates - use for starting custom drag operations, text selections, or long-press interactions. Must be followed by MouseUp action to complete.

Move mouse cursor to precise coordinates without clicking - perfect for triggering hover effects, revealing tooltips, and positioning for subsequent interactions.

Release mouse button at coordinates - completes drag operations, text selections, and long-press interactions. Should be used after MouseDown to finish mouse button sequences.

START HERE - Creates browser session and loads URL. Creates a browser with: - Stealth mode (anti-bot protection) - CAPTCHA solving (automatic) - Profile persistence (stays logged in) Always show debugUrl to user so they can watch the browser live. WORKFLOW: Navigate(url) -> FetchWebpage() -> MouseClick/PerformWebTask -> verify RETURNS: - sessionId: Required for all subsequent actions - debugUrl: Live browser view URL (share with user) - pageSnapshot: Initial page content FORCE NEW SESSION: If browser is stuck or in bad state, use forceNewSession=True to: - Kill the current session - Start fresh with a new browser instance

Paste text content at the current cursor position - perfect for filling forms, inserting data into text fields, or quick content insertion at focused elements.

AI agent for autonomous multi-step web interactions. CONSTRAINTS: - 50 step maximum (each click/type/scroll = 1 step) WHEN TO USE: - Multi-step workflows (login, fill form, submit) - Complex interactions where selectors are unknown - Dynamic content that requires visual understanding WHEN NOT TO USE: - Simple navigation: use Navigate - Reading page content: use FetchWebpage - Single click with known selector: use MouseClick GOOD PROMPTS (fit in 50 steps): - "Click login, enter email 'test@test.com', click submit" - "Search for 'laptops' and click first result" - "Fill form with name='John', email='john@test.com', submit" BAD PROMPTS (too many steps): - "Scrape all products from entire catalog" - "Log in, change all settings, download all reports" BREAK DOWN COMPLEX TASKS into multiple calls.

Capture high-quality screenshot of any webpage with extensive customization options - perfect for archiving, visual documentation, full-page captures, and cross-device viewport testing.

Scroll page to reveal hidden content. WHEN TO USE: - Element not visible on current viewport - Need to load infinite scroll content - Navigate long pages SCROLL AMOUNTS: - deltaY=200: Fine scroll (one section) - deltaY=400: Medium scroll - deltaY=800: Large jump - deltaY=-500: Scroll UP (negative) WORKFLOW: Scroll(deltaY=400) -> FetchWebpage() -> verify element visible

Store text content in the system clipboard for later paste operations - perfect for preparing data transfers, staging content for forms, or cross-application data sharing.

VISUAL VERIFICATION: Capture screenshot - MUST BE CALLED ALONE in multi_execute. 🚨 CRITICAL: Screenshot cannot be combined with other tools in same multi_execute call FALLBACK: If screenshot blocked/fails, use FetchWebpage diligently for information USE: Debug UI, verify states, document results after page changes RETURNS: S3 URL to downloadable screenshot image

Type text at current cursor position. PREREQUISITE: Click on input field first. WORKFLOW: 1. MouseClick(selector='input#email') 2. TypeText(text='hello@example.com') DELAY OPTIONS: - delay=0: Instant (default) - delay=50: Human-like speed - delay=100+: Slow, deliberate SPECIAL CHARACTERS: - Newlines: Include \n in text - Tab between fields: Use KeyboardShortcut(['Tab'])