Build AI agents that don't just answer questions—they take action. Take on the Google Cloud Rapid Agent Hackathon, where you'll move beyond the chatbot and into the world of agents that can reason, plan, and execute tasks.
This hackathon brings together Google Cloud and multiple technology partners—Arize, Elastic, GitLab, MongoDB, and Fivetran—to give your agents the tools they need to tackle challenges that matter.
In this post, we'll walk you through the key hackathon details, explore what each partner brings to the table, and show you how to get started.
Key details
- Submission period: May 5, 2026, to June 11, 2026
- Total prize amount: $50,000
- What to build: A functional AI agent—powered by Gemini and Google Cloud Agent Builder—that integrates one of the partners' MCP servers to solve a real-world challenge in your work, personal life, hobbies, or daily routines.
- What to submit: A URL to your hosted project, a URL to your public open-source code repository, a ~3-minute demo video, your selected partner track, and your completed Devpost submission form.
About the prize structure: This hackathon features multiple separate prize pools—one for each partner track. Rather than competing against every participant for a single prize, you'll compete within the category of the partner technology you chose. Each partner pool awards a first-place ($5,000), second-place ($3,000), and third-place ($2,000) winner.
Let's explore what this hackathon has in store and how to get started.
What you'll build
The Google Cloud Rapid Agent Hackathon is designed for builders at every level. Whether you're writing your first script or architecting complex systems, your mission is the same: engineer a functional agent that solves a real-world problem.
You'll use Gemini and Google Cloud Agent Builder as your build environment and a partner's MCP server as the set of tools that makes it all possible. Your agent shouldn't just answer questions; it should plan, act, and execute tasks across multiple steps—keeping you in control the whole time.
Here's what you'll do:
- Choose your partner: Pick the partner technology whose capabilities best match the agent you want to build.
- Set up your environment: Get your Google Cloud environment up and running with Gemini Enterprise Agent Platform and Agent Builder. Sign up for a no-cost trial at cloud.google.com/free
- Build your agent: Create an agent that uses tools to accomplish real tasks—managing data, automating workflows, or interacting with live services.
- Integrate your partner's MCP server: Connect your agent to your chosen partner's platform to unlock its superpowers.
- Submit your project: Share your hosted project URL, your open-source code repository, and a ~3-minute demo video showing your agent in action.
Now let's take a closer look at each partner and what they bring to the hackathon.
{{callout-1}}
Arize
Arize is an AI engineering platform built to help you develop AI apps and agents faster and perfect them in production. With Arize Phoenix, your Gemini-powered agent gets production-grade tracing, plus the ability to query its own traces, prompts, datasets, and experiments as tools at runtime via the Phoenix MCP server. Every decision it makes becomes inspectable, evaluable, and improvable.
How to build with Google Cloud and Arize
Build an agent that doesn't just run—it self-improves. Instrument your Gemini-powered agent with Arize Phoenix, then connect the Phoenix MCP server so your agent can introspect its own operational data at runtime. Strong submissions will demonstrate a meaningful self-improvement loop: the agent uses its own observability data to evaluate outputs and get better over time.
Some ideas to spark your thinking:
- An agent that traces its own reasoning steps and uses LLM-as-a-Judge evals to score and refine its responses
- A debugging assistant that queries its own trace history to identify patterns in failures and adjust its approach
- An agent that queries its own trace data to identify failure patterns and automatically adjusts its approach based on evaluation results
- A quality evaluation agent that runs automated evals on a dataset of past outputs and surfaces regressions
One important note: The Arize track requires a code-owned agent runtime such as Gemini CLI, the Gemini Enterprise Agent Platform SDK, Google ADK, Agent Runtime, or Cloud Run. The visual Agent Builder alone is not supported for tracing integration—you'll need to be able to instrument your code directly.
Arize resources
Getting started is fast: create a free Phoenix Cloud account, grab your API key, and you're tracing in under five minutes. Phoenix is also fully open-source if you prefer to self-host.
See the full list of resources and details on the Arize partner tab.
Elastic
Elastic, the Search AI Company, integrates its deep expertise in search technology with artificial intelligence to help everyone transform all of their data into answers, actions, and outcomes. Elastic's Search AI Platform — the foundation for its search, observability, and security solutions — is used by thousands of companies, including more than 50% of the Fortune 500.
Resources to help kick off your project with Elastic will be shared soon. In the meantime, keep an eye on the hackathon’s Updates tab.
Fivetran
The Fivetran platform moves, manages, and transforms data from every system a business runs on into a secure, reliable foundation with the flexibility to work across clouds, engines, and tools. With Fivetran, your analytics, operations, and AI run on data you can trust and control.
How to build with Google Cloud and Fivetran
There are two options for integrating your agents with the Fivetran platform. You can either build an agent that uses Fivetran's MCP server or REST API, depending on your use case and development style.
Some ideas to spark your thinking:
- An agent that monitors Fivetran pipeline health, detects sync issues, and takes corrective action
- A data discovery agent that helps users understand what data is available across connected sources and surfaces relevant tables or fields
- An agent that triggers pipeline runs based on external events and summarizes what data was loaded
- A business intelligence agent that draws on freshly synced Fivetran data to generate automated reports or summaries
Fivetran resources
Start with a free 14-day trial. Integrate your agent using either the Fivetran MCP server or the REST API. To use either option, you’ll need a Fivetran API key (the same key can be used for both REST and MCP).
See the full list of details and resources on the Fivetran tab.
GitLab
GitLab is a complete DevSecOps platform delivered as a single application that fundamentally changes how development, security, and operations teams collaborate to build software.
How to build with Google Cloud and GitLab
Take advantage of GitLab’s 30-day Ultimate trial to get access to everything you need to build an agent that uses the GitLab MCP server.
Some ideas to get you started:
- An agent that reviews open merge requests, summarizes changes, and flags potential issues for the author
- A release management agent that monitors pipeline status and surfaces blockers automatically
- An agent that triages incoming issues, classifies them by severity, and assigns them to the right team members
- A developer onboarding agent that walks new contributors through a codebase using GitLab's structure and history
Note: If you're using external tools to call GitLab via MCP, you'll need to set a default Duo namespace. The MCP server documentation covers this setup.
GitLab resources
The 30-day trial gives you access to everything you need—including Duo Agent Platform with 24 credits per user—with no access codes required.
MongoDB
MongoDB Atlas serves as the unified operational foundation and persistent memory layer for modern AI and agentic workloads. By combining operational, vector, and semantic data on a single platform, it eliminates the fragmented stacks and memory barriers that slow AI performance—letting you build production-grade agents that reason accurately across your data, regardless of your framework.
How to build with Google Cloud and MongoDB
Connect your Gemini-powered agent to MongoDB Atlas via the MongoDB MCP Server to give it a persistent, queryable data layer. Use Atlas Vector Search for semantic retrieval, Atlas Search for full-text queries, and aggregation pipelines to process and analyze data at scale—all within a single platform.
Not sure where to start? Use the sample Mfix Dataset to quickly spin up sample data for your project. The sample_mflix.embedded_movies already contains vector embeddings, so you can get started with Vector Search right away.
Some ideas to spark your thinking:
- An agent with persistent memory that recalls past interactions and uses them to personalize future responses
- A research agent that stores and semantically searches a personal knowledge base, surfacing relevant information on demand
- An agent that uses MongoDB Vector Search to power recommendations from a large catalog of items
- A data analysis agent that runs aggregation pipelines on operational data and delivers plain-language summaries
MongoDB resources
Get started: Steps to join the Google Cloud Rapid Agent Hackathon
Ready to start building? Here's how to get started:
- Register for the hackathon on Devpost.
- Choose your build environment. For a low-code option, start with Agent Builder. For custom agent logic, use the Gemini Enterprise Agent Platform SDK for Python. Use the Gemini Enterprise Agent Platform API Setup as the jumping-off point for either path.
- Get access to Google Cloud by signing up for a no-cost trial at cloud.google.com/free
- Choose your partner integration. Pick the MCP server whose capabilities best fit the agent you want to build.
- Start building! Submit your completed project before June 11.
Pro tip: Give yourself a leg up by submitting your project at least one week before the submission deadline. This gives the Devpost team a chance to review your submission and confirm it's eligible before the deadline. Check out all the benefits of submitting your project early.
Ready to build? Join the Google Cloud Rapid Agent Hackathon
The Google Cloud Rapid Agent Hackathon is your chance to go beyond the chatbot and build something that actually gets things done. With Gemini powering your agent's reasoning, Google Cloud Agent Builder as your build environment, and multiple partner MCP servers giving your agent real-world capabilities, the possibilities are wide open.
Register today to get started. We can't wait to see what you build!
