Skip to main content

NandaHack: Agentic AI Hackathon

HCLTech and MIT Media Lab

Credit Sponsors: OpenAI Codex

Build a system, tool, or layer that fixes a real bottleneck in how AI agents are deployed, coordinated, or scaled.

March 27 – May 7, 2026

5
Arenas
2
Tracks
Apr 11
Buildathon

What is the Agentic AI Hackathon?

A hackathon to build the missing infrastructure layer for AI agents

The Challenge

Six arenas. Escalating difficulty. Your agents must deploy, coordinate, trade, reason, and survive adversarial attacks. Not another chatbot. A battle-tested proving ground.

Why It Matters

The agent ecosystem is growing fast, but the infrastructure hasn't kept up. Coordination is ad-hoc. Memory is siloed. Communication is fragmented. We need builders who fix the plumbing.

What We're Looking For

Battle-tested agents that survive real pressure. Not slideware. Not vaporware. Agents that deploy, coordinate, trade, and defend under adversarial conditions.

Built on NEST

All agents are deployed on NEST infrastructure. Your solution should be compatible with and deployable on the NEST platform, ensuring real-world usability from day one.

About the Event

Solve core AI agent infrastructure problems. Build systems, tools, and layers that fix real bottlenecks in how AI agents are deployed, coordinated, and scaled.

The Challenge

Build agent infra so good that a generic AI can use it with zero custom integration

The “internet of agents” only works if infrastructure is good enough that any agent can use it without bespoke integration. That's exactly what we test. On demo day, a stock OpenClaw agent gets your instructions and tries to use what you built — live, in front of judges. No hand-holding. No custom glue code. It either works or it doesn't.

01

Build Infra

Teams build infrastructure projects for AI agents — wallets, registries, auth layers, coordination protocols, and more.

02

Demo Day

Judges spin up a stock, unmodified OpenClaw agent. No custom integrations. No special wrappers. Just a generic agent out of the box.

03

Plain English Test

The judge tells the agent in plain English to use your infra. If it can follow your instructions and interact with your project, it works. If it can't, it doesn't.

Live Demo — Pass / Fail

Can a generic agent actually use your infra from plain-English instructions alone? Concrete, objective, and on the spot.

Ambition & Vision

Subjective points for the scope of the problem you tackled and the boldness of your approach. Big swings matter.

Challenge Tracks

Two tracks. Choose one or combine across both.

Track 1

Pipeline Impact Mapping

Build AI solutions that directly map to enterprise pipeline and revenue generation.

TrackWhat It BuildsPipeline ImpactRevenue Lever
Client 0Internal AI agents, copilots, productivity toolsCreates proven use cases + credibilityFaster deal conversion (proof-led selling)
Enterprise AI (incl. Modernization)Client-ready solutions (SAP, legacy, agents)Generates direct qualified opportunitiesLarge transformation deals ($$$)
Sales AI Enablement (Client 0 x Enterprise)AI-powered account insights, RFP agentsImproves deal identification + win ratesPipeline expansion + higher conversion
Responsible AISafety, governance, red-teaming solutionsReduces enterprise adoption frictionUnlocks regulated industry deals (BFSI, healthcare)
Executive AI CoachingAI copilots for CXOs + coaching modelOpens C-suite access + strategic dealsHigh-value consulting + transformation programs
Track 2

Core AI Agent Infrastructure

Game of Thrones-style arena battles that stress-test real agent infrastructure. Six arenas, escalating difficulty, one grand champion.

🔥

The Forge

+

Deploy or die

Deploy a working agent. Speed and reliability are everything. First to ship a production-grade agent wins.

🌐

The Commons

+

No agent is an island

Agents share memory and coordinate to solve problems no single agent can. Build collective intelligence that actually works.

The Bazaar

+

Trust no one, verify everything

Agents trade services, verify identity, and settle micropayments. Build the economic layer for the agent economy.

🎯

The Gauntlet

+

Adapt or be forgotten

Agents face a rapid-fire series of unknown challenges. No preparation, no hints. Pure reasoning and adaptation under time pressure.

🛡️

The Wall

+

Last agent standing

Survive adversarial attacks while completing tasks. Your agents face prompt injection, resource exhaustion, and hostile actors.

👑

The Throne

+

Rule them all

Full integration challenge. Deploy, coordinate, trade, and defend. The grand champion masters every layer of the stack.

Deploy via

OpenClawJoin39

Judging

Impact 40% / Technical 40% / Simplicity 20%

Timeline

Phase 1: Mar 27 – Apr 11
Phase 2: Apr 11 – May 7

Timeline

Announcement. Buildathon. Build. Awards.

AnnouncementMar 27, 2026

Hackathon Launches

  • 01Hackathon officially announced
  • 02Form teams and choose an arena
  • 03Start exploring the 5 arenas
  • 04Prepare for the buildathon
BuildathonApr 11, 2026

In-Person Buildathon

  • 01In-person buildathon event
  • 02Kick off with fellow builders
  • 03Kick off your project with your team
  • 04Mentor office hours and support
  • 05Start building your prototype
Build PhaseApr 11 – May 7, 2026

Keep Building

  • 01Continue building after the buildathon
  • 02Polish and optimize your solution
  • 03Submit code, video demo, and slides
  • 04Final submissions due before May 7
AwardsMay 7, 2026

Award Ceremony

  • 01Live demos and presentations
  • 02Judging: Impact, Technical Depth, Simplicity
  • 03Winners announced
  • 04Celebration

How to Participate

What your submission needs to include

01

Choose an Arena

Pick one of the 5 arenas — from The Forge to The Throne — and prepare your agents for battle.

02

Build It

Write code. Deploy it. Make sure other developers can actually use what you build.

03

Submit

Upload your code, video demo, and slides before the deadline.

Requirements

  • Must be usable by other developers — not just a demo
  • Must run or be deployable — no slideware
  • Show clear improvement over the status quo
  • Compatible with NEST infrastructure
  • All agents deployed on NEST
  • Open source or clearly documented API

Judging Criteria

We value working software over fancy presentations

40%

Impact

Does this meaningfully solve a real problem? Would other developers actually use it?

40%

Technical Depth

Is the implementation solid? Does it demonstrate deep understanding of the problem space?

20%

Simplicity

Is the solution elegant and easy to adopt? Can someone get started in minutes, not hours?

FAQ