GroundUp

GroundUp: The AI financial agent that helps Gen Z stop surviving and start building.

Comment

Inspiration

For my inspiration, I thought back to how GenZ's ongoing struggles with student loans, rising rent, and entry-level salaries that barely cover the basics. Most financial tools were built for older, wealthier users with established careers and savings. That's when I realized I wanted to build something that actually speaks to where Gen Z is right now, not where they hope to be in 20 years. GroundUp was born from that frustration.

What it does

The GroundUp project heavily focuses on helping Gen Z users budget smarter, tackle debt, and build toward real financial goals through natural conversation. Instead of filling out spreadsheets, users simply talk to the agent — it extracts their financial situation, builds a structured plan, and stores everything persistently in MongoDB. The dashboard gives a live snapshot of their financial health score, spending breakdown, debt payoff countdown, and a personalized roadmap from their current situation to long-term stability.

How we built it

I built GroundUp using Google Cloud Agent Builder with Gemini powering the agent's reasoning and planning capabilities. MongoDB Atlas serves as the persistent data layer, connected through MongoDB's MCP server deployed on Cloud Run — giving the agent real memory across every session. The frontend is a React web app with a chat interface, interactive dashboard, and data visualizations built with Recharts, all deployed publicly via Cloud Run.

Challenges we ran into

Integrating the MongoDB MCP server with Google Cloud Agent Builder required careful configuration of Cloud Run permissions, Secret Manager, and service account roles before anything would talk to each other. Getting the agent to reliably trigger the right MongoDB tool calls based on natural language — rather than just responding conversationally — took significant prompt engineering and testing. Scoping the project tightly enough to ship all core features within the hackathon timeline was its own challenge.

Accomplishments that we're proud of

I'm honestly proud that the agent genuinely reasons through a user's full financial situation rather than just answering questions in isolation — it plans, prioritizes, and follows up proactively. The Financial Health Score and Broke to Stable Roadmap turned abstract financial data into something emotionally meaningful and immediately actionable. Building a full multi-page app with persistent agent memory, live data visualizations, and a working MCP integration within a week felt like a real accomplishment.

What we learned

This project deepened our understanding of how to architect agentic systems that go beyond chat — specifically how to design tool calls, manage conversation state, and give an agent reliable access to persistent data through MCP. I also learned how much prompt engineering matters when the agent needs to decide between responding conversationally versus taking a data action. Google Cloud Agent Builder made orchestration surprisingly approachable once the permissions and infrastructure were properly configured.

What's next for GroundUp

If this project continues, I certainly want to add bank account integration so the agent can automatically track real transactions rather than relying on manual input. A mobile app version is the natural next step given that Gen Z manages almost everything from their phones. Long term, GroundUp could expand into investment guidance and credit building — a full financial companion for every stage of the Gen Z money journey.

Built With

Share this project:

Updates