🔍 Read the full analysis: Discover Tool Retrieval Features In The Gemini API on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Google announced Tool Retrieval in preview for the Gemini Interactions API on Nov. 19, 2026. It lets developers keep tool names and descriptions available while loading full parameter schemas only when requested; Google has not published general savings figures.
Google announced Tool Retrieval in preview for the Gemini Interactions API on Nov. 19, as detailed in the original analysis, allowing developers to defer full function parameter schemas until a model requests a particular tool. The feature is intended for agents connected to large tool libraries, where including every schema in each request can add input tokens, according to Google.
Developers configure server-side retrieval by adding a tool_search tool and setting defer_loading: true for functions whose full schemas should initially be withheld. The model can still see deferred tools’ names and descriptions. When it asks for a tool by name, the API returns that tool’s schema, after which the model can make the function call within the same request, Google says.
The retrieval process appears in the interaction’s steps as a tool search call, a result containing the schema, and then the function call. The application handles the function and returns its result as usual. Developers can leave frequently used tools loaded while deferring less common ones. Google says server-side retrieval requires no index for developers to build or host.
The feature also supports tools exposed through remote MCP servers. Developers with their own tool registry, embedding index or permission filters can instead use client-side retrieval: their application searches its own system and returns matching schemas. Google says that approach requires an additional round trip for each search.
Introducing Tool Retrieval in the Gemini API
Google’s preview feature loads tool parameter schemas on demand, helping agents work with large tool libraries while keeping unused details out of the initial request.
What changes for agents
Tool names and descriptions remain visible; full parameter schemas load when requested.
Developers mark selected functions for deferred loading.
The model requests a tool by name and receives its schema.
The model calls it, and the application handles the result as usual.
Choose a retrieval approach
Server-side retrieval suits managed tool libraries; client-side retrieval supports custom search systems.
- Uses the tool_search tool.
- No index to build or host.
- Supports remote MCP server tools.
- Searches a custom registry or index.
- Can apply per-user permission filters.
- Adds a round trip for each search.
Understand the cost trade-off
Tool Retrieval has no separate charge; standard model token rates apply.
Skip unused schemas
Savings depend on whether schemas left unloaded outweigh schemas retrieved.
Search and loaded schemas
Search adds a small fixed block of input tokens; loaded schemas count as input for the rest of the conversation.
Managing Larger Gemini Tool Libraries
Tool libraries can contain many functions, although an individual task may use only one or a few. Sending every parameter schema with each request can increase the input context devoted to tools that are not used. By deferring those details, Tool Retrieval changes how tool definitions enter a conversation without removing the model’s access to the functions.
Google says the design is intended to reduce input token use, latency and tool-selection difficulty as libraries grow. Those are stated goals, not measured outcomes in the announcement. The feature’s practical value will depend on how many schemas remain unused and how often searches are needed.
Cost savings are not guaranteed. Google says there is no separate charge for retrieval, but each search adds a small fixed block of input tokens. A schema that is loaded is billed as input at the model’s standard rate for the rest of the conversation. Teams whose agents use most tools regularly may see less benefit than those with large libraries of infrequently used functions.
Products Worth Considering

Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation

Jackery Solar Generator 1000 V2 with 100W Solar Panel

Bose QuietComfort True Wireless Earbuds with Active Noise Cancellation, Bluetooth 5.3, USB-C Charging, and IPX4 Water Resistance
As an affiliate, we earn on qualifying purchases.
How Deferred Schemas Fit Function Calling
In conventional function calling, developers include tool definitions and their parameter schemas with a request. Google’s preview offers another arrangement for larger tool collections: keep a tool’s name and description available, but supply its detailed schema only after the model identifies it as relevant. The function still runs through the application’s ordinary execution flow once called.
Google says deferred schemas are compatible with prefix caching, because schemas are added later rather than at the start of the prompt. The company says this can leave the cached prefix unchanged across turns. The announcement lists support for Gemini 3.6 and later, including all Gemini 4 models, through the Interactions API.
““Tool Retrieval lets your agent keep its full tool library on standby and load parameter schemas only when the current task needs them.””
Savings and Preview Limits Remain Unclear
Google has not provided a general estimate for token savings, latency changes or tool-selection accuracy. It recommends that developers measure the effect on their own traffic, since results depend on library size, which tools are used and how requests are structured. No comparison baseline or independent evaluation was included in the announcement.
The feature remains in preview, and Google has not specified a general-availability date or detailed preview limitations. It is also not yet clear how retrieval performs across different tool libraries or whether the added search step affects response time for particular workloads. Those questions require testing rather than inference from the feature description.
Teams Can Test Their Own Workloads
Developers can consult Google’s setup documentation and test Tool Retrieval with supported models in the Interactions API. Google has invited users to discuss the feature and provide feedback. For teams considering adoption, the next practical step is to compare requests with and without deferred schemas under their own traffic.
That testing can track total input tokens, search frequency, latency and whether the model selects the appropriate tool. Google has not announced when the preview will end or when broader availability might begin, so those milestones remain open.
Key Questions
What does Gemini Tool Retrieval do?
It lets developers defer full tool parameter schemas and load them when the model requests a tool. Tool names and descriptions remain available before the full schema is retrieved.
How do developers enable server-side retrieval?
Google says developers add a tool_search tool and set defer_loading: true for functions whose schemas should be held back. The API then returns a requested schema for the model to use.
Does Tool Retrieval guarantee lower costs?
No. Google says retrieval has no separate charge, but searches add a small fixed block of input tokens, and loaded schemas are billed at standard input rates. Savings depend on how many schemas are never loaded.
Which Gemini models support the feature?
Google’s announcement lists Gemini 3.6 and later, including all Gemini 4 models, through the Interactions API.
When will Tool Retrieval leave preview?
Google has not announced a general-availability date. The feature is currently described as being in preview.
Primary source: Google AI · via ThorstenMeyerAI.com
