HackConnect

Inspiration

Hackathons are about building, but the first four hours are often lost to the “setup struggle.” Hackers spend excessive time searching for teammates with the right skill overlap and even more time initializing GitHub repositories, folder structures, and development environments.

We built HackConnect to eliminate this friction by automating the journey from solo hacker to collaborative team, allowing participants to start building immediately.

What it does

HackConnect is a dual-purpose platform designed to accelerate hackathon collaboration.

Intelligent Matching HackConnect analyzes user skills and project interests to recommend teammates with complementary expertise, forming balanced teams across frontend, backend, and design.

Agentic Workspace Once a team is formed, users enter a terminal-style collaborative workspace where specialized AI agents actively perform setup tasks:

Initializing GitHub repositories

Creating project folder structures

Launching shared Replit development environments

All of this happens with a single command, and updates appear instantly for every teammate.

How we built it

The Brain Powered by Gemini 1.5 Flash, we built a custom Agent Orchestrator that dynamically switches system instructions based on the selected specialist agent (Architect, Scrum Master, Designer).

Frontend

-React-based “Hacker Terminal” UI

-Neon terminal-inspired aesthetics

-Sliding panels for chat, commands, and agent responses

Backend

-Node.js and Express

-Octokit API for GitHub automation

-Replit Import API for environment setup

Database

-MongoDB

Gemini API and Google Agents

Stores vectorized user skills and live team chat memory To ensure high-quality team formation, HackConnect computes a compatibility score between users.

The matching score S between User A and User B is: S = α * (|A skills ∪ B skills| / (|A skills ∩ B skills| + 1)) + β * I shared interests Where: A skills and B skills represent each user’s skill set I shared represents shared project interests α and β weight skill diversity over similarity

This prioritizes complementary skill sets to form well-rounded teams.

Challenges we ran into

Our biggest challenge was transitioning from a traditional chatbot to a true agentic system. Initially, the AI only explained how to use GitHub instead of performing actions.

We solved this by implementing strict function calling on the backend, allowing the AI to recognize intent (such as “create a repo”) and trigger Octokit automation scripts instead of returning text responses.

We also had to manage real-time UI synchronization so that when a repository or environment is created, the update appears instantly for all teammates.

Accomplishments that we're proud of

-Reduced project setup time from 30 minutes to 30 seconds

-Built true agentic AI that interacts directly with developer tools

-Automated GitHub repository creation and Replit environment setup

-Designed a collaborative terminal-style interface

-Implemented a practical, skill-based team matching algorithm

What we learned

The future of collaboration is not just human-to-human, but human-to-agent. By giving AI the ability to interact directly with developer tools, we transformed it from a passive assistant into an active collaborator that removes friction and accelerates productivity.