DEV Community

Kourtney Lee
Kourtney Lee

Posted on

Never Miss an AI Hackathon Again: Building an Autonomous Discovery Agent with Notion & Brave MCP

#devchallenge #notionchallenge #mcp #ai

Notion MCP Challenge Submission 🧠

This is a submission for the Notion MCP Challenge

What I Built

HackathonSniper is an autonomous AI agent designed to eliminate "hackathon fatigue." If you're like me, you spend hours every month checking Devpost, DoraHacks, and Dev.to just to find one hackathon that fits your niche.

HackathonSniper turns this manual hunt into an automated pipeline. It doesn't just "search"—it thinks. It uses the Model Context Protocol (MCP) to bridge the gap between AI reasoning and the real web.

Video Demo

Show us the code

GitHub logo klee1611 / HackathonSniper

AI-powered hackathon discovery agent using Brave Search + Notion MCP

🎯 Hackathon Sniper

A sophisticated TypeScript agent that automatically discovers, evaluates, and tracks hackathons using AI-powered analysis. Built with clean architecture patterns, real MCP server integration, and complete environment separation.

✨ Features

  • 🔍 Intelligent Search - Real Brave Search MCP server integration
  • 🤖 AI Evaluation - Groq-powered hackathon qualification analysis
  • 📊 Notion Integration - Real Notion MCP server for storage and tracking
  • 🎭 Test Isolation - Complete environment isolation for testing (zero external dependencies)
  • ⚡ Fast Development - Hot reload, comprehensive Makefile, pnpm package management
  • 🛡️ Type Safety - End-to-end TypeScript with Zod schema validation
  • 🧪 Robust Testing - Unit, integration, and smoke tests

🚀 Quick Start

Prerequisites

  • Node.js 18+
  • pnpm (recommended) or npm

Installation & Setup

# Clone and setup in one command
git clone <repository-url>
cd HackathonSniper
make setup
Enter fullscreen mode Exit fullscreen mode

Development Workflow

# Copy and fill in your API keys
cp .env.example .env
# Run development
…
Enter fullscreen mode Exit fullscreen mode
View on GitHub

The Technical Architecture

The agent follows a rigorous 3-phase lifecycle:

  1. 🔍 The Hunt (Brave Search MCP): It performs targeted searches across the web to find new hackathon announcements, even those buried in blog posts or obscure landing pages.
  2. 🤖 The Vetting (Groq LLM): Every discovery is analyzed by a Llama-3 model. It evaluates the prize pool, solo-friendliness, AI relevance, and proximity to the deadline.
  3. 📊 The Storage (Notion MCP): Qualified opportunities are instantly synced to a beautifully structured Notion database, complete with metadata, deadlines, and AI-generated summaries.

tech_architecture

How I Used Notion MCP

The Notion MCP Server (@notionhq/notion-mcp-server) is the "nervous system" of this project. It transforms a simple script into a powerful productivity tool by allowing the AI to interact with my Notion workspace as if it were a native application.

Instead of writing complex, boilerplate-heavy API calls to the Notion REST API, the Notion MCP is integrated. This allowed the agent to:

  • Dynamic Schema Mapping: The agent automatically maps structured AI evaluation objects to Notion database properties (Dates, Checkboxes, URLs, and Rich Text).
  • Process Isolation: By using the MCP StdioClientTransport, the Notion integration runs in its own isolated process, managed entirely by the Model Context Protocol.
  • Automated Workflows: The agent uses the create_notion_page tool provided by the MCP server to dynamically populate my "Hackathon Scout" dashboard.

Wrapping Up

This integration unlocks a "set and forget" workflow. I can run HackathonSniper as a background task, and my Notion workspace stays updated with a high-signal, low-noise list of opportunities. It’s not just data entry; it’s an autonomous data clerk powered by Notion MCP.

Top comments (0)