Google adds tool retrieval feature to Gemini API agents

Covered by 2 sources
Read onTLDR AI· 6 min
Google launched a Tool Retrieval feature in its Gemini Interactions API that lets agents dynamically load and execute functions from large tool libraries, improving efficiency and reducing latency. The feature supports both server- and client-side retrieval to enhance flexibility and accuracy in tool selection.

What happened

On November 19, 2026, Google launched a Tool Retrieval feature in its Gemini Interactions API. This update allows agents using the API to dynamically load and execute functions from large libraries of tools as needed. The feature is available in both server-side and client-side retrieval modes, giving developers more options for integrating and managing tools. The goal is to make workflows more efficient by reducing the need to load all tool information up front.

Key details

  • Tool Retrieval was introduced in the Gemini Interactions API preview on November 19, 2026.
  • Agents can now execute registered functions automatically based on input, such as issuing refunds.
  • The feature supports both server-side and client-side retrieval modes.
  • Tool schemas are only loaded when needed, which reduces token usage and latency.
  • Developers can register functions for the API to locate and execute automatically.

Why it matters

Developers using the Gemini API can now manage large tool libraries more efficiently, which can lower infrastructure costs and speed up application response times. This change allows companies and teams building AI-powered applications to offer more flexible and accurate automation without increasing latency or resource use. The new feature also gives developers more control over how and when tools are accessed, which can improve the reliability and scalability of their applications.

Also covered by 1 source


More in Dev