On July 25, 2026, Amrita Vishwa Vidyapeetham, Coimbatore turned into a 24-hour building floor.
No pause after the workshop. No take-home assignment. No week-long submission window.
The hackathon clock started at 1:30 PM on July 25 and stopped at 1:30 PM the next day.
More than 350 teams registered, over 220 teams actively participated, and Amrita reported 177 final project submissions across six tracks.
For NitroStack, this was exactly the kind of Agentic AI ecosystem we want to build: give students the infrastructure, make MCP understandable, and then let them see how far they can take it when the clock is actually running.
First, make MCP understandable
Before students started building, the event opened with conversations around Agentic AI, entrepreneurship and the growing connection between industry and engineering education.
Abhishek Pandit, CEO of NitroStack, delivered a keynote during the inaugural session.
Then the NitroStack technical team, led in part by Luis Felipe Hernandez Mora, took students into the practical layer: what the Model Context Protocol (MCP) is, why agents need it, and how to actually build with it.
That distinction matters.
Explaining that MCP connects AI systems with tools and data is easy enough.
Giving hundreds of students a stack where they can actually turn that idea into a working application is much harder.
That's the gap NitroStack is built to close.
Then 24 hours of nonstop MCP building
Once the hackathon began, teams worked across six areas:
- BFSI & FinTech
- Education & Research
- Enterprise AI & Workplace Automation
- HealthTech & Life Sciences
- Manufacturing & Industry 4.0
- Open Innovation
The projects became technically ambitious very quickly.
Ekalavya 2.0, the Education & Research winner, was built as a complete MCP server using the NitroStack TypeScript SDK, deployed through NitroCloud and tested with NitroStudio.
The team built 35+ tools across six autonomous agent modules, covering areas such as skill analysis, research, project generation, career guidance and resume development.
That's a fairly serious agentic architecture to ship in one day.
Another winner, ForgeOps, tackled explainable multi-agent decision-making for manufacturing.
ScamShield approached financial fraud and scam detection through AI workflows.
NutriPulse applied agentic systems to healthcare and nutrition.
And those were only a few of the projects.
Across the hackathon, students built systems around infrastructure remediation, cybersecurity, research, government services, manufacturing, financial workflows, healthcare and enterprise automation.
NitroStack made the infrastructure disappear
This is one of the reasons these university hackathons matter so much to us.
If every student had first needed to understand container orchestration, manually wire an MCP runtime, configure deployment infrastructure, build authentication from scratch and separately assemble debugging tools, most of those 24 hours would have disappeared into plumbing.
Instead, NitroStack gives builders an end-to-end MCP development path.
The open-source NitroStack TypeScript SDK provides the framework for tools, resources, prompts, authentication, middleware and application logic.
NitroStudio lets students build, inspect and test what their MCP server is doing.
NitroCloud gives them a path to deploy the application rather than leaving it running on a laptop.
So students could focus on the actual product:
Problem → Agent architecture → MCP tools → Test → Deploy → Demo
That is a much better use of 24 hours.
Students shouldn't have to wait to become AI infrastructure engineers
One assumption around technologies such as MCP is that students need years of backend or infrastructure experience before they can work with them properly.
Coimbatore showed something different.
Give students an approachable MCP stack and a real problem, and they start building surprisingly sophisticated systems very quickly.
That's why NitroStack is taking MCP so seriously at the university level.
The next generation of AI developers shouldn't enter the industry knowing only how to prompt a model.
They should understand how agents connect to systems.
How tools are exposed.
How permissions work.
How applications are deployed.
And how an AI system moves from generating an answer to actually taking an action.
That is the infrastructure layer behind Agentic AI.
This is how an ecosystem gets built
The hackathon ended with winners across all six tracks receiving cash prizes, NitroStack credits and continued opportunities to develop their projects. Winning teams were also offered incubation support through Amrita's AICACE ecosystem.
But the bigger result was the volume of actual building.
Hundreds of teams.
A full overnight sprint.
Production-style MCP applications.
And students who, within a single event, went from understanding MCP to shipping with it.
For NitroStack, that's the objective.
Don't just spread awareness of Model Context Protocol. Put the infrastructure in developers' hands and create the next generation of MCP builders.
At Amrita University Coimbatore, that happened for 24 hours straight.