Skip to Content
User Guide🎁 Special FeaturesSmart Tool Discovery

Smart Tool Discovery

Smart Tool Discovery helps you automatically search for and execute tools that match the user’s prompt intent. By automatically finding and running the necessary MCP tools in AI Hub, you can use LLMs more efficiently and register multiple tools even in environments with tool count limitations.

Key Benefits

  • Improved Response Efficiency → Instead of passing the entire tool list to the LLM, only the two tools needed for searching and execution are provided, reducing token consumption. When many tools are registered in a preset, this also helps the LLM decide which tool to call, improving tool invocation performance.

  • Flexible in Environments with Tool Count Limits → Some LLM applications (such as Cursor IDE, Claude desktop app, etc.) limit the number of tools you can register. However, Smart Tool Discovery dynamically searches for and provides tools, allowing you to bypass these restrictions.

Feature Overview

Smart Tool Discovery consists of two main components:

  • Search Tool: Finds the most appropriate tool for the user’s prompt and informs the LLM.
  • Execution Tool: Executes the tool selected by the LLM and returns the result.

How It Works

Example Scenario

  1. The user asks in natural language, “Summarize today’s schedule.”
  2. The Search Tool searches the preset tool list and recommends relevant tools.
  3. The Execution Tool runs the selected tool and returns the result.

How to Use

  1. Activate 40 or more tools in a preset.
  2. The Smart Tool Discovery feature will be automatically enabled, and a popup will notify you. Alternatively, go to Settings → Labs → toggle Use Smart Tool Discovery to enable or disable the feature.

Summary

  • Automatically searches for and executes necessary tools based on natural language requests
  • Reduces the number of tools passed to the LLM, optimizing token usage
  • Works smoothly even in environments with tool count limitations

Smart Tool Discovery automates the process of natural language request → relevant tool search → tool execution, making integration with AI-based systems even easier.

Last updated on
💬 Supported LLM AppsCustom MCP Server