GCP CloudMate AI

• 

  System architecture

Inspiration

Today’s developers, startups, and even enterprises are expected to adopt cloud technology at a rapid pace—but navigating platforms like Google Cloud can be daunting. Between budgeting, selecting the right services, understanding architectures, and provisioning resources, the learning curve is steep.

I am inspired by the potential of Agentic AI to make cloud computing more intuitive and autonomous. With tools like the Agent Development Kit (ADK), I see a unique opportunity to create a natural language interface that simplifies the cloud journey—turning complex tasks into effortless conversations.

What it does

GCP CloudMate AI is a multi-agent system that allows users to interact with Google Cloud through natural language.

With just a sentence like “I want to build an e-commerce app,” the system can:

• Recommend the right GCP services based on user needs and budget
• Generate detailed system architecture
• Create or manage cloud resources like Firestore databases or Storage buckets

Let your ideas meet the cloud—instantly, intelligently, and effortlessly manage it—with GCP CloudMate AI.

How we built it

The project is powered by the Google Cloud Agent Development Kit (ADK) and is composed of four collaborating agents:

1. gcp_advisor_agent – Interprets user goals and budget, and recommends suitable GCP services
2. architecture_agent – Translates system requirements into a high-level architecture design
3. gcp_management_agent – Manages GCP resources via natural language commands (e.g., create/delete buckets, Firestore DBs)
4. orchestrator_agent – Routes user requests to the right agent based on intent

Tech Stack:

• Python for backend logic
• Gemini models for language understanding and task reasoning
• Streamlit for UI
• Cloud Run for deployment and scalability

Challenges we ran into

• Agent Discoverability: As the system grew, it became difficult for the orchestrator to always determine which agent should handle ambiguous or multi-intent queries.
• Custom Tool Development: Many GCP services aren’t MCP-compliant, so I had to manually implement SDK-based tools for Firestore and Cloud Storage to work within the ADK ecosystem.
• Natural Language Mapping: Interpreting varied user input accurately, especially vague or incomplete instructions, required prompt engineering and iteration.

Accomplishments that we're proud of

• Built a fully working multi-agent system capable of real-time collaboration across AI agents
• Created a functional, user-friendly interface that allows natural interaction with Google Cloud
• Successfully provisioned and managed real GCP resources using plain text input
• Demonstrated the power of ADK and Agentic AI in building real-world autonomous assistants

What we learned

• How to use Agent Development Kit (ADK) to design interoperable agents
• The power and flexibility of Agent-to-Agent (A2A) communication
• How to integrate LLMs and Google Cloud SDKs for automation
• A deeper appreciation for the potential of Agentic AI to revolutionize cloud workflows

What's next for GCP CloudMate AI

• Expand GCP capabilities: Add support for more GCP services like Cloud Run, Pub/Sub, BigQuery, and IAM
• Improve Orchestration: Implement an intelligent routing system with better context tracking
• Deploy a chatbot interface: Bring CloudMate AI to messaging platforms and developer tools
• Open Source It: Share the project with the community to inspire and build together

Thank you for exploring GCP CloudMate AI. The future of cloud management is conversational—and it’s just getting started.

Built With

• cloud
• firestorre
• gcp
• python
• storage