ActivePage returns the currently active page.

RefreshElementMap updates the cached element map.

WaitStable waits for the page to become stable.

Get returns an element by index.

Click clicks on an element by index.

Capture takes a screenshot of the page and returns compressed bytes. It implements proper page readiness checks following browser-use best practices.

Center returns the center point of the bounding box.

DefaultAnnotationConfig returns sensible defaults for annotations.

GetElementMap returns the current element map.

Navigate opens a URL in the browser. This is a convenience method for direct navigation without a task.

highlightElement shows a visual highlight on an element.

humanDelay adds a random human-like delay.

Close shuts down the browser and cleans up resources.

LLMOptions returns options optimized for LLM vision consumption. Includes page readiness checks and content validation.

ToTokenStringLimited is a convenience method with a max elements limit.

clamp restricts a value to be within min and max.

Annotate draws bounding boxes and labels on a screenshot image. Following browser-use pattern: boxes around interactive elements with index labels.

BuildPageStatePrompt creates a prompt describing the current page state.

CaptureWithAnnotations captures a screenshot and draws element annotations. This follows browser-use pattern: bounding boxes around interactive elemen

DefaultAnnotatedOptions returns options with annotations enabled.

Description returns a human-readable description of the element.

NewElementMapAdapter creates an adapter for annotation.

NewExtractor creates a new DOM extractor.

Run executes a task described in natural language. Returns a Result containing the outcome and execution details.

Screenshot takes a screenshot of the current page. Uses the enhanced screenshot package with proper page readiness checks.

ScreenshotSafe takes a screenshot, returning nil (not error) for blank pages. This is useful for agent loops where blank screenshots should be skipped

ScreenshotSafeWithAnnotations takes an annotated screenshot, returning nil for blank pages.

Start launches the browser and initializes the agent.

containsIgnoreCase checks if s contains substr (case-insensitive).

isBlankPage checks if the page is a blank page (about:blank or empty).

Add adds an element to the map.

Contains checks if a point is within the bounding box.

Extract extracts interactive elements from the page.

FilterSensitiveData filters sensitive data from a message.

GetConsecutiveFailures returns the number of consecutive failures at the end.

GetElementMap extracts interactive elements from the current page.

IsPageReady checks if the page is ready for screenshot capture.

ListTabs returns information about all open tabs.

New creates a new browser automation agent. Call Start() before using Run().

New creates a new browser instance.

NewTab opens a new browser tab.

ScreenshotAfterAction captures a screenshot after an action completes. Waits for page stability before capturing.

StepCount returns the number of steps executed.

applyStealthMode injects stealth JavaScript into a page.

captureAndSaveScreenshot captures a screenshot and saves it to disk if configured. Returns the screenshot bytes and the saved path (empty if not saved

clearInput clears the currently focused input.

createMultimodalContent creates a genai.Content with both text and image.

generateTabID creates a unique 4-character tab ID.

isEmptyImage checks if the image is mostly white/empty. Uses sampling to check a grid of pixels for efficiency.

randomMouseOffset returns a small random offset for more human-like clicking.

toLower converts ASCII characters to lowercase.

AddHistoryItem adds an item to the execution history.

AddItem adds a new history item.

BuildContinuationMessage builds a message for continuing after an action.

BuildErrorRecoveryPrompt creates a prompt for recovering from an error.

BuildInitialTaskMessage builds the initial message with task description.

BuildStateMessage builds the current state message for the LLM.

BuildTaskPrompt creates the initial task prompt.

Clear resets the history.

ClearAndType clears an input and types new text.

Close shuts down the browser and cleans up resources.

CloseTab closes a tab by ID.

CreateAllTools creates all browser automation tools.

CreateClearAndTypeTool creates the clear_and_type function tool.

CreateClickTool creates the click function tool.

CreateCloseTabTool creates the close_tab function tool.

CreateDoneTool creates the done function tool.

CreateDoubleClickTool creates the double_click function tool.

CreateEvaluateJSTool creates the evaluate_js function tool.

CreateExtractContentTool creates the extract_content function tool.

CreateFocusTool creates the focus function tool.

CreateGetPageStateTool creates the get_page_state function tool.

CreateGoBackTool creates the go_back function tool.

CreateGoForwardTool creates the go_forward function tool.

CreateHoverTool creates the hover function tool.

CreateListTabsTool creates the list_tabs function tool.

---- Tool Functions ---- CreateNavigateTool creates the navigate function tool.

CreateNewTabTool creates the new_tab function tool.

CreateReloadTool creates the reload function tool.

CreateScreenshotTool creates the screenshot function tool.

CreateScrollToElementTool creates the scroll_to_element function tool.

CreateScrollTool creates the scroll function tool.

CreateSendKeysTool creates the send_keys function tool.

CreateSwitchTabTool creates the switch_tab function tool.

CreateTypeTextTool creates the type_text function tool.

CreateWaitTool creates the wait function tool.

DefaultSerializeOptions returns sensible defaults.

DefaultStealthConfig returns sensible stealth defaults.

DefaultViewport returns the default viewport size.

DoubleClick double-clicks on an element by index.

EstimateTokens estimates the number of tokens in a text string. This uses a simple character-based heuristic: - ~4 characters per token for English te

EvaluateJS evaluates JavaScript code on the page.

ExtractContent extracts text content from the page.

Filter replaces sensitive data with placeholders.

Focus focuses on an element.