Global AI Hackathon - Devcon 2026 By Mel
Global AI Hackathon presented by mel
Online · Saturday, October 31, 8:00 ET to Sunday, November 1, 2:00 PM ET
Build something real with Mel, the agentic terminal that turns plain English into working projects.
The challenge: Agents on Duty
Don't build an app. Hire an agent.
Pick one track and one real customer: a friend's shop, your team at work, a local business, a club, a nonprofit. Then build a team of AI agents in Mel that does a real job for them. Your agents read, decide, act and check their own work, while a human stays in charge.
The examples below are there to get you started. Adapt them, or solve a very specific problem for your own customer.
Headline track: Black October Emergency Patch (Agentic Engineering)
Why now: October 2026 brought a wave of critical security flaws: PHP2Shell (CVE-2026-109618 and CVE-2026-109619), which CISA lists as exploited in the wild, and HTTP2Shell, a flaw in the Brotli compression library (CVE-2026-111497). Teams everywhere are racing to patch.
Example: a three-agent team. A security agent reads server logs and tells a real attack marker from a harmless error. An engineer agent traces it to the vulnerable version and writes the patch. A QA agent runs the tests. Before anything deploys, the team stops and asks a human.
Rules: work only in your own safe test setup: a small test app pinned to an old version, with fake logs. No real attack code, and no servers you don't own.
Score: minutes from the first alert to a tested patch.
AI Employees: the Tribal Knowledge Agent
Why now: 29% of companies that cut staff for AI have hired people back, because the know-how left with them (Robert Half).
Example: an internal agent that answers "how do we do this here?" questions, such as "how do we deploy?" or "how do we set up a new service?", by reading a team's real docs, past tickets and code.
Must: cite the exact source for every answer, and hand off to a human when it isn't sure.
Enterprise AI: Legacy Code Rescue
Why now: TCS and Workato launched agentic legacy modernization on October 27, 2026. Big companies are paying to move old code.
Example: take one service from a small open-source legacy app (old Java, .NET, PHP or COBOL). Your agents read it, plan the move, rewrite it and run it in a container.
Must: prove with tests run in the terminal that the old and new versions give the same output.
Agentic Commerce: Ready for AI Shoppers
Why now: Adobe reports that AI traffic to US retail sites grew 393% in Q1 2026 and converts 42 to 54% better, yet the average homepage is only 75% readable by AI.
Example: make a store readable and buyable for AI shopping agents. Your agents clean the product data, fix the store pages, and then a shopping agent completes a test checkout from start to finish.
Must: use a free Shopify development store with its test payment gateway, or your own store with Stripe test mode. Test payments only, and the owner approves before any change goes live.
Agentic Banking, Finance & Payments: the Reconciliation Agent
Example: an agent that compares a company's ledger (a spreadsheet or CSV) with payments in a sandbox such as Stripe test mode. It finds the mismatches, fixes the small ones and explains every number in plain words.
Must: fake data only, and a human approves before any money moves. No tax or investment advice.
Agentic Trading: the AI Infrastructure Shock Desk
Why now: Amazon raised its 2026 capital spending to $220 billion, citing the cost of memory chips. Chips, energy and supply chains are moving markets.
Example: an agent that tracks public news and data on chips, energy and supply, re-tests a paper trading strategy, and proposes changes with a backtest and risk notes.
Must: paper trading only (for example, a free Alpaca paper account). A human makes every decision.
Agentic Marketing: the Campaign Planner
Example: connect Mel to an open-source social media scheduler such as Postiz or Mixpost. Your agents take one product announcement, plan a schedule across channels, write the posts for each platform, build a landing page and queue everything.
Must: a human reviews every post and page in Mel before anything is scheduled or published.
Agentic Revenue: the Pipeline Builder
Example: an agent that builds a qualified lead list from public company data, drafts personal outreach for each lead, and keeps a pipeline dashboard up to date.
Must: public data only (respect each site's terms), and it drafts only. It never sends by itself.
Wildcard: your own idea
Build anything useful with a team of agents in Mel. Same rules as everyone else.
Rules for every project
It runs live, on a public link or a real machine, working on real or realistic data.
It checks its own work: tests, a linter or a verification step after each action.
It knows its limits. It handles something going wrong and hands off to a human when unsure.
A human approves before any money moves, any code goes to production, or anything posts.
A real person tries it. Show their reaction, with their permission.
Event schedule (Eastern Time)
Saturday, 8:00 AM ET: kickoff. Start building. Your Mel hackathon credits are added after you register and sign up in Mel with the same email.
Sunday, 8:00 AM ET: the curveball. Every track gets one new request from its "customer".
Sunday, 11:00 AM to 2:00 PM ET: live session. Show your build, get mentor feedback, ask submission questions, polish your demo.
Sunday, 2:00 PM ET: final submission deadline on Devpost.
Sunday, 2:00 to 8:00 PM ET: judging. US clocks fall back early Sunday morning, so Sunday times are EST.
Build solo or in a team of up to 5. Bring an idea, ship a working demo, and show how Mel helped.
What judges look for
35% Usefulness: does it solve a real problem?
25% Execution: does the working flow feel complete?
20% Creativity: is the approach fresh and thoughtful?
20% Mel use: did Mel materially help planning, building, debugging, reviewing or testing?
Prizes
Track winners and finalists: 3 months of free building in Mel (credits worth up to $300)
Winners featured on Agentic Day and Mel channels
Access to the Mel builder community
Who it's for
Developers who want to see how far an agent can really go
founders and solo builders shipping fast
Product people experimenting with AI-native workflows
Non-technical builders who want to turn ideas into working software
No coding experience required.
What you need
A laptop and an internet connection
Mel installed: download at openmel.dev. We'll help with setup at kickoff.
Spots are limited. Register now and we'll send you the event link, community access, the challenge brief and your Mel starter kit.
Join Mel's WhatsApp group · Join Mel's Discord
Stuck? Post in WhatsApp or Discord, or email hello@openmel.dev. Don't spend hours stuck alone.
Pick one real problem, build the smallest version that works, and show it. Go build.
Mel x Agentic Day
Would you like to co-host with us? Reach out at lalla@helloagentic.ai.