Skip to content

WillStark/Magnetron

Repository files navigation

Magnetron

Catch buying intent before it shows up in search.

A B2B lead-discovery agent that crawls Reddit, Hacker News, and Twitter for active buying signals — complaints about competitors, "best X for Y" questions, problem descriptions matching your ICP — then enriches each signal with company + decision-maker context and a drafted outreach DM.

Built for the AI Agent Economy Hackathon (AgentHansa, 2026-04-25).


The pitch (3 sentences)

Most lead-gen tools find people who might buy. Magnetron finds people who are actively unhappy with your competitor and ready to switch. Point it at your ICP and 2–3 competitor names, and 90 seconds later it surfaces 20 ranked leads with sentiment-weighted scores, source links, and personalized DMs ready to send.

Why this wins

  • Real demand on AgentHansa: ResearchBot (the top A⭐ showcase agent) earns money doing exactly this shape of work — we 10x it and add an enrichment + outreach layer.
  • Sponsor-aligned: TokenRouter ($200 demo credit) powers the enrichment + DM-drafting pipeline.
  • Self-monetizing demo: agent literally claims live AgentHansa quests during the 3-min pitch. Slide says "$X earned tonight" with proof.
  • $500/mo story: one closed deal pays for a year of service.

Repo layout (planned)

Magnetron/
├── README.md           # this file
├── PLAN.md             # locked-in build plan, architecture, demo flow
├── RESEARCH.md         # AgentHansa platform intel (what's saturated, what sells)
├── CODEX_REVIEW.md     # second-opinion feedback from codex
├── .env.local          # TokenRouter key (NEVER COMMIT)
├── .env.local.example  # template for sharing
├── .gitignore          # Next.js + secrets
└── (Next.js app to be scaffolded)

Quick start (once scaffolded)

yarn create next-app@latest . --typescript --tailwind --app --use-yarn
yarn dev

Stack

  • Next.js 16 + Tailwind 4 (single-page app)
  • Reddit JSON API + HN Algolia API (free, public)
  • TokenRouter (api.tokenrouter.com/v1) — enrichment + DM drafting
  • DeepSeek-chat — fallback scoring model
  • AgentHansa expert API — register agent + claim quests

Submission checklist

  • Public GitHub repo
  • 2–3 min demo video (MP4/WebM/MOV up to 500 MB, or Loom/YouTube/Drive link)
  • AgentHansa expert profile (bonus)
  • Submit at agenthansa.com/hackathon-submit.html before 8:00 PM PST

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors