TartanHacks 2026 Recap: From Workshop to Working Agents in 24 Hours
February 10, 2026 by Parteek Singh
1,042 people checked in to TartanHacks 2026, making it the biggest edition yet of Pittsburgh's largest hackathon. Organized by ScottyLabs at Carnegie Mellon, TartanHacks is a 24-hour sprint that produced 269 projects this year. The theme was "Mosaic": teams assemble scattered fragments of vision, skill, and code into something greater than any single part.
Agentuity was proud to be a sponsor at the event, making this our second time at CMU after ScottyLabs' Nova GenAI Hackathon last November. Nova was seven hours and focused on GenAI. TartanHacks was 24 hours, with a much broader scope: fintech, gaming, accessibility, creative tools, and more.
Before the Clock Started
The Monday before hacking began, we ran a workshop introducing Agentuity v1 to participants. We walked through the platform's core features, including sandboxes, storage services, and the OpenCode plugin (Agentuity Coder), and ran a live demo using a hackathon starter project.
By Friday evening, hacking was underway.
The Hackathon Floor
Alongside sponsors like AppLovin, Capital One, Ripple, and CodeRabbit, there were dozens of prize tracks across a wide range of categories.
Students were curious about the kinds of agents that developers and customers are actually building using Agentuity, and asked about features like evals and the AI Gateway. In general, people were happy to talk agents. We offered free credits to participants, and a number of students signed up to try the platform out.
Projects That Caught Our Eye
After 24 hours of hacking, teams lined up to present at the judging expo.
Some of the most interesting projects we saw at TartanHacks:
One team built a video game designed to get people moving. It scans your body into the game as a playable character and uses ML to track metrics like speed and movement in real time, turning exercise into gameplay.
Another team built a mind-mapping app where you speak your thoughts instead of typing them. The app captures voice, generates visual mind-maps, and lets you share in collaborative spaces. The idea is that talking through something is more raw and authentic than writing it down, and the dashboard they built was well done.
But the project we liked most (and admittedly, we're a little biased!) was the one that used Agentuity.
Ledgerly: AI-Powered Financial Planning with Agentuity
One team built Ledgerly, a real-time financial planner and budgeting app that connects to your bank accounts and uses AI agents to deliver proactive insights on spending and budgeting. They used Agentuity to power the agent layer.
Their financial-insights agent runs a three-step pipeline: first it reasons about a transaction in the context of spending history, budgets, and financial goals. Then it builds a plan with goal-oriented action items. Finally, it decides whether to notify the user, with severity levels (info, warning, or alert). Notifications get persisted to a key-value store and surface in the app through the @agentuity/react provider.
The result is a financial coaching app that watches spending patterns and nudges users when something looks off, rather than waiting for them to check a dashboard.
We asked the team what it was like building with Agentuity during the hackathon:
"Building Ledgerly AI with Agentuity was incredibly efficient. Agentuity's runtime made it easy to implement complex behaviors like our 'Financial Insights' agent. We could define clear perception, reasoning, and planning stages using standard schemas, and the framework handled the orchestration perfectly. It turned what could have been a messy web of API calls into a structured, reliable, and scalable agent architecture."
Check out the team's full project: Agentuity agents, backend, and frontend.
Acknowledgments
Two trips to CMU now, and each one has shown us something new about how students are thinking about agents. If you were at TartanHacks and want to keep building on Agentuity, the docs are the best place to start. If you're an AI coding agent, we wrote an onboarding guide just for you.
Huge thanks to the ScottyLabs team for putting together a great event, to the Ledgerly team for building with our platform, and to everyone at CMU who stopped by our booth.
We hope to see you again, CMU!