Google Cloud revives AI agent roadshow for Cloud Run

Google Cloud has revived its Accelerate AI with Cloud Run roadshow for 2026, focusing on building and running AI agents on Cloud Run.

The roadshow returns with an updated curriculum centred on what Google describes as the full lifecycle of AI agents, from early deployment to operational use in business settings. The workshop uses a fictional "Coffee Shop Journey" to guide participants through a series of increasingly complex tasks.

Google is targeting developers and practitioners working on agent-based AI applications rather than simple prototypes. The programme is built around the gap between quickly assembling an AI tool and turning it into a service that can operate at scale.

Participants begin by deploying a simple web application on Cloud Run in a coffee shop launch scenario to learn the basics of the platform. The sessions then move to building a coffee recommendation agent with Google's Agent Development Kit and retrieval-augmented generation, before expanding into analytics and workplace assistance tools.

Another part of the curriculum uses Gemma 4 and the BigQuery MCP server to analyse bike routes and identify locations for hypothetical coffee stands. A separate exercise focuses on a productivity assistant for store managers, using Cloud Run to support day-to-day operational tasks and scheduling.

Agent lifecycle

The programme also introduces Antigravity 2.0, which Google says can be used with skills, context, rules and hooks to add features to Cloud Run applications. This year's emphasis is on the later stages of development, where teams need to manage orchestration, long-running agents and data-driven decision-making rather than simply produce an initial demo.

That marks a shift from the rapid experimentation that has defined much of the recent AI software boom. Google wants Cloud Run to be seen not just as a place to deploy lightweight applications, but as an environment for the ongoing operation of agentic systems that depend on a mix of models, tools and data services.

The workshops highlight several Google products together. Alongside Cloud Run, Google points to its Agent Development Kit, BigQuery MCP and Gemma 4 models, while also highlighting GPU support on Cloud Run for low-latency model inference without requiring users to manage clusters directly.

Cloud strategy

For Google, the roadshow is also a way to tie its AI software and cloud infrastructure messages more closely together. The curriculum links model use, retrieval workflows, application hosting and data analysis into one developer path, reflecting how large cloud providers are trying to keep AI development within their own platforms.

Cloud Run has traditionally been positioned as a managed, serverless environment for running containers and web services. By putting AI agents at the centre of this training effort, Google is signalling that it sees the service as relevant to newer application patterns involving persistent workflows, tool use and model interaction, rather than only short-lived request handling.

The workshop structure also suggests Google wants to appeal to teams looking for practical examples tied to business use cases. The coffee shop theme serves as a teaching device, but the underlying tasks map to common enterprise concerns, including recommendation systems, location planning, staff productivity and integrating generative AI into everyday operations.

In the source material, Google drew a distinction between quickly building agents with tools such as Antigravity and AI Studio and the work required to move those systems into routine use. "While 'vibe coding' with tools like Antigravity and AI Studio lets you build and deploy complex agents in minutes, the real work begins on 'Day 2', " said Google Cloud.

It also explained the revised focus of the programme: "This year, we've updated our curriculum to focus on the full AI agent lifecycle, giving you the keys to productionizing and scaling agentic workloads on Google Cloud's serverless platform," said Google Cloud.

The workshops are intended to give attendees direct experience of building, scaling and orchestrating long-running AI agents while combining model access, analytics and application deployment on Cloud Run. The curriculum is designed around "real-world business problems as you evolve from launching a simple cafe to building complex, intelligent assistants."