@hyperbrowser/agent

import { Browser, BrowserContext, Page } from "playwright-core"; import { BrowserProviders, HyperAgentConfig, MCPConfig, MCPServerConfig } from "../types/config"; import { ActionType, Task, TaskOutput, TaskParams } from "../types"; import { HyperPage, HyperVariable } from "../types/agent/types"; export declare class HyperAgent<T extends BrowserProviders = "Local"> { private static readonly AIACTION_CONFIG; private llm; private tasks; private tokenLimit; private debug; private mcpClient; private browserProvider; private browserProviderType; private actions; private cdpActionsEnabled; browser: Browser | null; context: BrowserContext | null; private _currentPage; private _variables; private errorEmitter; get currentPage(): HyperPage | null; set currentPage(page: Page); constructor(params?: HyperAgentConfig<T>); /** * This is just exposed as a utility function. You don't need to call it explicitly. * @returns A reference to the current rebrowser-playwright browser instance. */ initBrowser(): Promise<Browser>; /** * Use this function instead of accessing this.actions directly. * This function configures if there is a need for an output schema as a part of the complete action. * @param outputSchema * @returns */ private getActions; /** * Get all variables * @returns Record of variables */ getVariables(): Record<string, HyperVariable>; /** * Set a variable * @param key Key of the variable * @param value Value of the variable */ addVariable(variable: HyperVariable): void; /** * Get a variable * @param key Key of the variable * @returns Value of the variable */ getVariable(key: string): HyperVariable | undefined; /** * Delete a variable * @param key Key of the variable */ deleteVariable(key: string): void; /** * Get all pages in the context * @returns Array of HyperPage objects */ getPages(): Promise<HyperPage[]>; /** * Create a new page in the context * @returns HyperPage object */ newPage(): Promise<HyperPage>; /** * Close the agent and all associated resources */ closeAgent(): Promise<void>; /** * Get the current page or create a new one if none exists * @returns The current page */ getCurrentPage(): Promise<Page>; /** * Get task control object for a specific task * @param taskId ID of the task * @returns Task control object */ private getTaskControl; /** * Execute a task asynchronously and return a Task control object * @param task The task to execute * @param params Optional parameters for the task * @param initPage Optional page to use for the task * @returns A promise that resolves to a Task control object for managing the running task */ executeTaskAsync(task: string, params?: TaskParams, initPage?: Page): Promise<Task>; /** * Execute a task and wait for completion * @param task The task to execute * @param params Optional parameters for the task * @param initPage Optional page to use for the task * @returns A promise that resolves to the task output */ executeTask(task: string, params?: TaskParams, initPage?: Page): Promise<TaskOutput>; /** * Find element with retry logic * Retries element finding with DOM refetch until element is found or max retries reached * * @param instruction Natural language instruction for the action * @param page The page to search on * @param maxRetries Maximum number of retry attempts * @param retryDelayMs Delay between retries in milliseconds * @returns Object containing the found element, DOM state, and element map * @throws Error if element is not found after all retries */ private findElementWithRetry; private writeDebugData; /** * Collect interactive elements from element map for debugging * Extracts elements with interactive roles (button, link, textbox, etc.) * * @param elementMap Map of element IDs to element data * @param limit Maximum number of elements to collect * @returns Array of interactive elements with id, role, and label */ private collectInteractiveElements; /** * Execute a single granular action using a11y mode * Internal method used by page.aiAction() * * Architecture: Simple examine->act flow * - 1 LLM call (examineDom finds element and suggests method) * - Direct execution (no agent loop) * * @param instruction Natural language instruction for a single action * @param page The page to execute the action on * @returns A promise that resolves to the task output */ private executeSingleAction; /** * Register a new action with the agent * @param action The action to register */ private registerAction; /** * Initialize the MCP client with the given configuration * @param config The MCP configuration */ initializeMCPClient(config: MCPConfig): Promise<void>; /** * Connect to an MCP server at runtime * @param serverConfig Configuration for the MCP server * @returns Server ID if connection was successful */ connectToMCPServer(serverConfig: MCPServerConfig): Promise<string | null>; /** * Disconnect from a specific MCP server * @param serverId ID of the server to disconnect from * @returns Boolean indicating if the disconnection was successful */ disconnectFromMCPServer(serverId: string): boolean; /** * Check if a specific MCP server is connected * @param serverId ID of the server to check * @returns Boolean indicating if the server is connected */ isMCPServerConnected(serverId: string): boolean; /** * Get all connected MCP server IDs * @returns Array of server IDs */ getMCPServerIds(): string[]; /** * Get information about all connected MCP servers * @returns Array of server information objects or null if no MCP client is initialized */ getMCPServerInfo(): Array<{ id: string; toolCount: number; toolNames: string[]; }> | null; /** * Pretty print an action * @param action The action to print * @returns Formatted string representation of the action */ pprintAction(action: ActionType): string; getSession(): Browser | import("@hyperbrowser/sdk/types").SessionDetail | null; private setupHyperPage; }