🔍 Read the full analysis: Build With Tool Retrieval In The Gemini API on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Google announced Tool Retrieval in preview for the Gemini Interactions API on Nov. 19, 2026. It lets developers defer full function parameter schemas until a model requests a tool, while keeping names and descriptions available; Google has not published general savings figures.
Google announced Tool Retrieval in preview for the Gemini Interactions API on Nov. 19, 2026, giving developers a way to keep full function parameter schemas out of a model request until a tool is needed. The change is aimed at agents connected to large tool libraries, where sending every schema can add input tokens and make tool selection more difficult, according to the original analysis.
To use the server-side feature, developers add a tool_search tool and set defer_loading: true for functions whose schemas should be held back. The model can still see each deferred function’s name and description, but not its full parameter schema. When it requests a function by name, the API loads the schema, allowing the model to make the function call within the same request. For related Gemini API capabilities, see Gemini Decisions API.
Google says the interaction steps show the sequence: a tool-search call, a search result containing the schema, and then the function call. The application handles that function and returns its result as usual. Developers can leave commonly used tools loaded in advance and defer less frequently used ones. Google says server-side retrieval requires no index for the developer to build or host.
The feature also supports remote MCP servers: setting deferred loading on an MCP server tool makes its exposed tools available on demand. Developers with their own registry, embedding index or per-user permission filter can use client-side retrieval instead. In that setup, the application searches its own system and returns matching schemas, but each search adds a round trip.
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 Retrieval addresses a practical scaling issue for agents that can call dozens or hundreds of functions. In a conventional setup, developers include tool definitions and their parameter schemas in a request. That can add context even when a task uses only one function. Deferring schemas gives teams a way to expose a broad library without automatically sending all of its parameter details at the start.
Google says the design is intended to reduce input-token use, latency and difficulty selecting tools as libraries grow. Those are stated goals, not published benchmark results. The cost calculation is workload-dependent: Google says there is no separate fee for retrieval, but the search adds a small fixed block of input tokens, and each schema loaded is billed as input at the model’s standard rate for the rest of the conversation.
That means savings depend on how many schemas remain unused. If an agent loads most of its tools frequently, deferral may save little and the retrieval step may add overhead. Teams will need to compare total usage and request behavior against their own traffic rather than assume that a larger library automatically brings lower costs.
Products Worth Considering

Samsung Galaxy Tab A11+ Kids Edition, 128GB, Royal Blue

Kindle Paperwhite Signature Edition 32GB (2024 Model) – Metallic Black

Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
As an affiliate, we earn on qualifying purchases.
How Deferred Schemas Fit Function Calling
In standard function calling, a developer supplies tool definitions, including parameter schemas, with a request. Google’s announced approach separates the information the model sees initially: names and descriptions remain available, while full schemas are supplied only after the model identifies a relevant function. The function still runs through the application’s usual handling and result-return process.
Google says Tool Retrieval works with Gemini 3.6 and later, including all Gemini 4 models, through the Interactions API. The announcement also says deferred schemas are compatible with prefix caching: because schemas are added later rather than at the beginning of the prompt, Google says the cached prefix can remain consistent across turns. The feature is in preview, and the company has invited developers to consult its setup documentation and provide feedback.
““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 Unknown
Google has not provided a general estimate for token savings, latency changes or tool-selection accuracy. It advises developers to measure the effect on their own traffic. Results may vary with the size of a tool library, how often functions are used and whether server-side or client-side retrieval is chosen.
The announcement does not include independent evaluations of retrieval accuracy as libraries grow, nor does it spell out all preview limitations. It also gives no date for general availability. The available information establishes how Google says the feature works, but not whether it improves cost or speed for a particular production workload.
Measure Results During the Preview
Developers can test Tool Retrieval with supported models through the Gemini Interactions API while the feature remains in preview. A useful evaluation would compare requests with and without deferred schemas on representative traffic, tracking total input tokens, latency, retrieval steps and whether the model selects the appropriate function.
Google has invited feedback, but has not announced a general-availability schedule or a broader performance report. For now, teams considering the feature will need to consult the setup documentation, check how their chosen model and tools behave, and decide from their own measurements whether deferral reduces overhead.
Key Questions
What is Tool Retrieval in the Gemini API?
Tool Retrieval is a preview feature in the Gemini Interactions API that lets developers defer full function parameter schemas until a model requests a particular tool. Tool names and descriptions remain visible beforehand.
How do developers enable server-side Tool Retrieval?
Google’s described setup adds a tool_search tool and sets defer_loading: true for functions whose schemas should be loaded on demand. The API returns the requested schema before the model makes the function call.
Does Tool Retrieval reduce Gemini API costs?
It may reduce input-token use when many schemas are never loaded, but Google has not published a general savings figure. The search adds a small fixed block of input tokens, and loaded schemas are billed at the model’s standard input rate.
Which Gemini models support the feature?
Google says Tool Retrieval works with Gemini 3.6 and later, including all Gemini 4 models, through the Interactions API.
When will Tool Retrieval be generally available?
Google announced the feature in preview on Nov. 19, 2026, and has not provided a general-availability date.
Primary source: Google AI · via ThorstenMeyerAI.com
