import { AgentExecutor } from 'langchain/agents';

/**
 * Custom AgentExecutor that intercepts large tool outputs and converts them to content references
 * before they are sent to the LLM to avoid token limit issues.
 *
 * Note: The content reference conversion is already handled in the MCP adapter,
 * so this class currently just extends AgentExecutor without modifications.
 * We keep it as a placeholder for future enhancements.
 */
export declare class ContentAwareAgentExecutor extends AgentExecutor {
    private logger;
    constructor(config: any);
}