Google AI has launched the Gemini API Tool Retrieval feature
The story so far
Google has launched the Tool Retrieval feature in its Gemini API, allowing agents to dynamically load tool parameter schemas to reduce the number of input tokens and maintain the accuracy of tool selection. This feature is implemented by adding the tool_search tool in the request and setting defer_loading: true, and is applicable to Gemini 3.6 and later models, including all Gemini 4 models, with no additional cost. Using tool_search adds a small fixed number of input tokens, and the schemas loaded by each model are billed based on input, the same as pre-declared tools.
AI-written from the reports · Updated 4 hours ago
Coverage timeline
Follow the coverage to see every side of the story.
- Google AI BlogIntroducing Tool Retrieval in the Gemini API
The Gemini API now introduces Tool Retrieval in preview, allowing agents to dynamically load tool parameter patterns when needed, reducing the number of input tokens while maintaining the accuracy of tool selection. This feature is implemented by adding the tool_search tool in the request and setting defer_loading: true, supporting Gemini 3.6 and higher models, including all Gemini 4 models, at no additional cost.
Heat over time
- Now
- 9
- Peak
- 920 Nov, 04:00
- 24h change
- –
Hover or tap the chart to see each hour; use the arrow keys on a keyboard.