Tool Retrieval in the Gemini API
In brief
Gemini's Interactions API now supports on-demand loading of tool parameter schemas. Agents with large tool libraries can keep schemas out of context until needed, cutting input tokens, latency, and selection errors.

How it works
Add a tool_search tool to your Interactions API request and set defer_loading to true on function or remote MCP server tools. The model sees only tool names and descriptions upfront, then fetches full schemas inside the same API call when needed. For custom registries or per-user filtering, you can set execution to client to run the search in your own app and return schemas in a follow-up turn. It supports Gemini 3.6 and later models.
Developers building Gemini agents that connect to large tool libraries or MCP servers.
Availability
Sources
11 merged into this card- 1Introducing Tool Retrieval in the Gemini APIAnnouncement · blog.google · 19.11
- 2Choosing a modelHow-to guide · aistudio.google.com
- 3Function calling with the Gemini APIHow-to guide · aistudio.google.com
- 4Interactive world generation with Genie 4How-to guide · aistudio.google.com
- 5Genie 4 prompt and controls guideHow-to guide · aistudio.google.com
- 6Interactions APIHow-to guide · aistudio.google.com
- 7Latest Gemini modelsHow-to guide · aistudio.google.com
- 8Genie 4How-to guide · aistudio.google.com
- 9Rate limitsHow-to guide · aistudio.google.com
- 10Tool retrievalHow-to guide · aistudio.google.com
- 11Gemini API release notes — November 19, 2026Release notes · aistudio.google.com · 19.11
More from Google

Computer use in Gemini app for Mac
Google added computer use to the Gemini app for Mac, letting Gemini automate multi-step tasks across local files, desktop apps, and the web in an isolated background window.
Gemini CLI /stats heatmap subcommand
Google added a /stats heatmap subcommand to Gemini CLI v0.69.0 to track daily token usage. It lets developers monitor their token consumption directly from the command line.