Round 01ONLINE· 2nd week · Nov 2026
Idea Round
Submit your idea online. Open to students from any college in India, in teams of 2 to 4.
ALLof India can enter
The Event · Code Relay 5.0
An online idea round open to all of India, then a 30-hour build sprint on campus at IIT Bhubaneswar. Here's everything you need.
✦Teams of 2–4✦Round 01 is free✦Open to all of India✦30-hour final✦100–150 teams on campus✦Web2 × Web3 · 4 tracks✦Teams of 2–4✦Round 01 is free✦Open to all of India✦30-hour final✦100–150 teams on campus✦Web2 × Web3 · 4 tracks
The formatHow the
How the
relay runs
Two rounds, one baton. Start online from anywhere in India, finish with a 30-hour sprint on campus. Follow the path.
Baton PassONLINE· After Round 01
The Shortlist
The strongest ideas get the baton. 100 to 150 teams are invited to the finale on campus.
100–150teams shortlisted
Round 02OFFLINE· 3rd week · Jan 2027
The Final Sprint
Shortlisted teams build a working prototype in a 30-hour hackathon at IIT Bhubaneswar.
30 HRSnon-stop on campus
FINISH
LINE
LINE
Winners take
the podium.
Choose your lane
Pick a domain, then one track inside it. Build agents and AI for Bharat in Web2, or onchain payments and AI × Web3 in Web3.
AI that takes over real work, and AI that works for every Indian.
Web2 · Track 01
with TrueFoundryAI & Automation
Agents that take over real work, and still ask before anything they can't undo.
Sub-tracks01
Autonomous Business Operations
Build agents on TrueForge that run a real business workflow end to end, such as a support resolution agent that reads order and billing data, issues refunds and escalates when it should, or a multi-agent workflow where separate agents handle procurement, dispatch and payment follow-up. Anything irreversible must pass through Tool Approval.
02
Engineering & Incident Response
Build agents that keep systems running. Examples are an incident response agent that reads alerts, logs and metrics, finds the likely cause of an outage and suggests a fix for a human to approve, or a DevOps agent that triages failing builds overnight and drafts fixes. Handling large logs without overflowing context is central, so Context Engineering and Sub-Agents matter here.
03
Data & Decision Automation
Build agents that turn raw data into decisions. Try a data analyst agent that answers business questions in plain language by querying databases and producing charts, or an inventory and logistics agent that predicts stock-outs, optimizes routes and drafts orders for approval. Memory should show how the agent improves from one run to the next.
04
Internal Operations & Workforce Automation
Build agents that remove routine work inside organizations. Examples include an internal helpdesk agent handling IT and HR requests with the right permissions per user, or a finance back-office agent that checks invoices, flags mismatches and sends reminders. Keep permissions tight, keep every action logged and make the agent say clearly what it could not do.
Web2 · Track 02
with CSM Technologies × TrueFoundryAI for Bharat
AI that fixes a real problem for the last user in the line, in India's languages, networks and scale.
Sub-tracks01
Governance & Public Services
Build AI that makes digital governance work for every citizen. Examples include an agent that takes grievances in Indian languages, routes them across departments and shows live status until closure, or a scholarship and certificate assistant that verifies documents from other systems instead of asking for uploads. Protect citizen data by design, work on a basic phone, leave an audit trail officers and citizens can follow, and use human approval before any action that cannot be undone.
02
Healthcare, Education & Rural Livelihoods
Use AI to close access gaps in health, learning, agriculture and climate resilience. Try a health worker assistant that guides referrals and flags drug shortages, a voice-first tutor for low-connectivity regions, or a farmer alert and registry desk that sends pest and weather warnings by SMS or voice. Prioritize local languages and simple interfaces, and hand over to a human professional whenever the AI is unsure.
03
FinTech & Financial Inclusion
Build secure financial services for people the system still misses. Ideas include a first-time credit guide, a fraud triage agent that groups likely scams and asks a human to approve any freeze or customer message, a micro-insurance claims desk, or a UPI scam-warning companion in the user's language. Explain every decision plainly, and keep a human in control of anything that moves money.
04
Mobility, Trade & Infrastructure
Build agents and tools that keep India moving and its businesses running. Examples are an incident response agent that reads traffic and sensor feeds and suggests diversions for approval, an EV charging and bus-reliability planner, a small-seller order and inventory desk, or a multi-agent trade workflow covering procurement, dispatch and payment follow-up. Handle large data feeds cleanly, show real-time tracking and give users dashboards they can act on quickly.
✦
+ Bonus track
Open Innovation
Build anything. If it's bold and it works, it belongs here.
✸
+ For every track
Partner Bounties
Sponsors run their own bounties, open to teams from every track.
Round 01 · Guidelines
Round 1 is free to enter and needs no paid tools or API keys: use any provider with a free tier, an open-source model or a local model. Only teams selected for Round 2 pay a registration fee; on-campus accommodation is optional.
AI & AutomationWeb2 domain · Track 01 · with TrueFoundry
Every organization has work that nobody wants to do by hand: triaging alerts, reconciling spreadsheets, answering routine requests, chasing approvals. This track asks you to build an AI agent that takes over real work, runs through a multi-step job on its own and still asks a human before it does anything it cannot undo.
What to buildDesign or build an AI agent that completes a real job of at least 5 steps across at least 2 tools or data sources. It must plan, act, recover from at least one failure, pause for human approval before any irreversible action and end with a clear report a human can act on in under 2 minutes. The track is built around TrueForge, TrueFoundry's open-source (MIT-licensed) agent harness. You are not required to use it: teams may build on TrueForge or any other framework, and no marks are lost either way.
Sub-tracks · choose one
- 01Autonomous Business Operations
- 02Engineering & Incident Response
- 03Data & Decision Automation
- 04Internal Operations & Workforce Automation
Agent design patterns
Use at least 4 of the 6 patterns; Human Approval is compulsory. For each one, state where in your workflow it is used and why the agent fails without it. A name-drop scores zero. Every pattern can be built with free tools.
01
Human Approval
A list of the irreversible actions and the exact point where the agent pauses for a human.
Compulsory02
Sub-Agents
How the job is split, with each sub-agent's role and what it hands back.
03
Memory
What the agent remembers between runs and how that changes the next run.
04
Context Management
How large outputs such as logs and tables are summarized or trimmed without overflowing the model's context.
05
Tool Use through MCP or APIs
Which tools the agent calls, why each is needed and what happens when a tool fails.
06
Long-run Execution
How a multi-step job is triggered and tracked, and how the agent resumes after an error.
Also declare the model(s) and why (cost, speed, privacy), the tools used, and the interface (chat UI, SDK, API or dashboard).
How you are judged
30
Agent design depth
20
Problem and agent fit
20
Safety and failure design
15
Architecture and feasibility
10
Report and user experience
5
Clarity of submission
Out of 100, plus a prototype bonus of up to 10 · Tie-breaker: the higher safety and failure design score.
Submission
Blueprint tier
Pitch deck (PDF, max 10 slides) including a Failure Drill slide, and a one-page Agent Card.
Prototype tier
The same, plus a demo video of up to 3 minutes showing one live human approval and the final report, and a public GitHub repo with a README.
Not accepted
Generic Q&A chatbots, “chat with your PDF”, plain RAG over notes, resume screeners and study buddies, unless they clearly meet every requirement. A plain chatbot wrapper with an agent name on it will be marked down. Never commit API keys; use synthetic or anonymized data only.
Round 02 · On campus
16:00 - 17:30
Registration & Check-In
Participant badging, mentor room setup, and technical environment checks.
17:30 - 19:00
Opening Ceremony & Keynote
Welcome address by WebnD IIT Bhubaneswar and industry track sponsors.
19:00 - 19:45
Problem Statement Reveal
Unveiling challenge repositories, evaluation metrics, and test datasets.
20:00
Hacking Marathon Begins (T=0)
Relay Sprint 1 officially kicks off. Teams commence ideation and architecture.
How we judge
Your demo is scored across four weighted criteria.
30%
Technical Depth & Complexity
Robustness of system architecture, concurrency handling, smart contract security, gas optimization, or distributed data flow.
25%
Innovation & Problem Solving
Novelty of the solution, distinct architectural choices, and practical value applied to the problem statement.
25%
Implementation & Working Demo
Completeness of the codebase, passing unit/integration tests, operational live deployment, and realistic reliability under load.
20%
UI/UX & Presentation
Intuitive user interface, coherent interaction design, accessibility, and the team's ability to clearly articulate their solution.
Who can play
01TEAM DYNAMICS
2–4 Hackers Per Team
Compete in collaborative cells of 2 to 4 members. Inter-college, inter-branch, and inter-year teams are fully permitted and welcomed.
02STUDENT STATUS
All Indian Colleges
Open to all active undergraduate (B.Tech, B.Sc, BCA, etc.) and postgraduate (M.Tech, MCA, etc.) students with a valid college ID card.
0330H MARATHON
Continuous Relay Sprint
30 uninterrupted hours of architecture, development, and debugging with milestone checkpoints and mentorship touchpoints.
04HYBRID VENUE
Online, Then On Campus
Round 1 is an online idea round you can enter from anywhere in India. Shortlisted teams build the final prototype on campus at IIT Bhubaneswar.
Loot drop
Hover or tap a coin to flip it.
WEB2 TRACK WINNER₹80,000++ CLOUD & API CREDITS
WEB2 TRACK WINNER
Awarded to the most innovative, scalable, and production-ready distributed system or intelligent web application in the Web2 track.
PRIZES & CERTIFICATES FOR WINNERSWEB3 TRACK WINNER₹60,000++ PROTOCOL GRANTS & AUDITS
WEB3 TRACK WINNER
Awarded to the top decentralized protocol, zero-knowledge application, or smart account infrastructure project in the Web3 track.
PRIZES & CERTIFICATES FOR WINNERSTOTAL PRIZE POOL₹2,50,000+INCLUDING IN-KIND REWARDS
TOTAL PRIZE POOL
Cash prizes for both track winners, plus runners-up awards and in-kind rewards from our partners.
PRIZES & CERTIFICATES FOR WINNERSFAQ
Everything you need before you grab the baton.
Join the participants' Discord at discord.gg/y54wHSpVC. Announcements, shortlist news, Round 2 details and answers from the organisers are all posted there first.
Any undergraduate or postgraduate student currently enrolled in a recognized university or institute in India is eligible. Teams may consist of 2 to 4 members. Inter-college and inter-branch teams are fully permitted.
Round 1 is free for everyone. Only teams selected for Round 2, the 30-hour final on campus at IIT Bhubaneswar, pay a registration fee to confirm their place. Accommodation on campus is optional and charged separately. Selected teams get the exact amounts along with their shortlist confirmation.
Participants can register online for round 1 and submit their ideas. If qualified for round 2, they will have to attend it on-campus at the IIT Bhubaneswar permanent campus (Argul, Jatni).
Code Relay 5.0 has two domains with two tracks each. Web2: AI & Automation (with TrueFoundry) and AI for Bharat (with CSM Technologies × TrueFoundry), each with four sub-tracks. Web3: Onchain Payments and AI × Web3, each with a problem statement and starting ideas, open to any blockchain. Each team picks one track.
All code submitted for judging must be written during the 30-hour hackathon duration. You may utilize public open-source libraries, packages, frameworks, and APIs, but the core project logic, contracts, and interfaces must be original work created during the event.
The Code Relay 5.0 tracks, sub-tracks and Round 1 guidelines are live on the Event page and in the PS Vault. The vault also has the problem statements from Code Relay 4.0 and 3.0.
Code Relay 5.0 · #SprintToInnovation
Readysetgo
Round 01 is online and open to every college in India. 2–4 per team. Round 01 is free.