Google DeepMind upgrades Gemini API with multi-tool chaining and Google Maps as data source
Google DeepMind has shipped a significant upgrade to the Gemini API, giving developers more powerful primitives for building agentic workflows.
What's new?
The headline feature is multi-tool chaining — developers can now combine built-in tools like Google Search and Google Maps with custom functions in a single API request. Previously, each step had to be handled separately, making workflows slower and more cumbersome to build.
Context circulation
Results from one tool can now be automatically passed to the next through what Google calls "context circulation." This enables smooth, multi-step AI workflows without manual hand-offs between calls. Each tool call also gets a unique ID, making debugging significantly easier.
Google Maps as a data source
Another key addition: Google Maps is now available as a data source for the Gemini 3 model family, providing direct access to location data, business information, and commute times within AI workflows.
Google recommends its new Interactions API for building these workflows.
What does this mean for developers?
The update is a clear move toward more autonomous AI agents. Instead of firing separate requests for each step, Gemini can now orchestrate complex tasks in a single round — reducing latency and simplifying system architecture.
For companies building agentic applications on top of Gemini, this means a lower technical barrier and faster iteration cycles.
Competitive context
The update comes as all major AI players invest heavily in agent infrastructure. Anthropic recently made 1M token context windows generally available, while OpenAI is building out Codex for coding agents. Google's answer is clear: better tool integration and richer data sources baked directly into the API.
Combined with Gemini 3.1 Pro's position as Google's most capable publicly available model as of March 2026, this strengthens Google's bid to be the infrastructure platform of choice for AI agent builders.
📬 Likte du denne?
AI-nyheter for ledere. Kuratert av en CIO som bygger det selv. Daglig i innboksen.