Google Gemini API Adds Context Circulation and Multi-Tool Chaining — Agents Can Now Combine Google Maps, Search, and Custom Functions in Single Requests

On March 19, 2026, Google DeepMind announced significant upgrades to the Gemini API tooling stack, introducing capabilities that transform how AI agents orchestrate multi-tool workflows. These updates make Gemini 3 one of the most capable model APIs for agentic multi-tool scenarios.

CONTEXT CIRCULATION:

The headline feature is context circulation — the ability for results from one tool call to be automatically passed as context to subsequent tool calls within a single API request. Previously, developers had to manually handle the intermediate results, creating orchestration overhead. Now, the model can chain tool calls naturally, with each step enriched by the outputs of previous steps.

For example, an agent can:

1. Search for a product using a custom API → get result
2. Look up the store location on Google Maps → get coordinates
3. Check real-time availability using another API → combine all context
4. Generate a response that synthesizes all information

All within a single API request with no developer-side orchestration code.

MULTI-TOOL COMBINATIONS:

Built-in tools (Google Search, Google Maps grounding) can now be freely combined with custom developer-defined functions in a single request. This removes a major limitation — previously, developers had to choose between built-in tools OR custom functions, not both.

The Decoder reports: Google has created an SDK and MCP server for Stitch, so users can link it to coding assistants, extending the multi-tool paradigm across design and development workflows.

GOOGLE MAPS GROUNDING:

A new grounding capability for Google Maps enables location-aware agent workflows. Agents can now ground their responses in real geographic data — store locations, routes, nearby points of interest, and real-time traffic information. This is particularly powerful for:

• E-commerce agents providing store pickup options
• Travel planning agents with real route calculations
• Local service agents recommending nearby businesses
• Logistics agents optimizing delivery routes

DEVELOPER EXPERIENCE:

Dr. Susan Thomas from Google published the announcement emphasizing that these updates target complex agentic workflows: whether combining Grounding with Google Maps with internal inventory APIs, or letting the model naturally circulate context between custom tools, developers can turn complex agentic workflows into reality easily.

The updates are available immediately for Gemini 3 Pro and Gemini 3 Flash models via the Gemini API, Google AI Studio, and Vertex AI.

COMPETITIVE CONTEXT:

This positions Gemini API as the most feature-complete model API for multi-tool agent scenarios, ahead of:

• OpenAI function calling (no built-in tool combinations)
• Anthropic tool use (no native Maps/Search grounding)
• Mistral tool calling (no context circulation)

The updates come the same day as Google Stitch vibe design platform launch with MCP integration, suggesting a coordinated push to make Google the dominant platform for agent-first development workflows.