Skip to main content
MP
← Back to Projects
AI Agents · B2B Sales · Automation2026

HUNTR

“Your AI sales team. Without the payroll.”

Role

Solo Developer

Stack
Google ADKGemini 2.5 ProGemini 2.5 FlashPythonFastAPINext.jsFirestoreCloud RunVercel
Status

Live

Overview

HuntR is an autonomous B2B client acquisition agent that deploys 5 specialized AI agents to find, research, score, and personally message your next client — in under 2 minutes.

I built this because most Indian startups lose deals not because their product is bad, but because their outreach is generic, late, and manual. A junior sales rep costs ₹840+ per lead in time. HuntR brings that to zero.

Built end-to-end solo in under 6 hours for Agentathon 2026 — from blank repo to production-deployed multi-agent system. Uses Google ADK for orchestration with a dual-model strategy: Gemini 2.5 Pro for reasoning-heavy agents and Gemini 2.5 Flash for speed-optimized ones.

Screenshots

Key Features

→
Scout Agent — Finds 20+ leads showing real pain signals across LinkedIn, Reddit and IndiaMART
→
Researcher Agent — Enriches each lead with company data, decision maker name and email
→
Scorer Agent — Ranks leads 1-100 based on fit, urgency and India-specific signals
→
Outreach Agent — Writes hyper-personalized cold email and LinkedIn message per lead using Gemini 2.5 Pro
→
Followup Agent — Builds Day 3, 7, 14 follow-up sequence automatically
→
Self-Correction — Manager Agent retries with refined queries when results fall below threshold
→
Email Open Tracking — Pixel-based tracking shows when decision maker opens your email in real time
→
Campaign History — Firestore-backed persistence with campaign comparison across hunts

Tech Deep Dive

I used Google ADK for multi-agent orchestration with a deliberate dual-model strategy — Gemini 2.5 Pro powers the Manager and Outreach agents where reasoning quality and personalization matter most, while Gemini 2.5 Flash handles Scout, Researcher, Scorer and Followup where speed is the priority. The FastAPI backend streams agent trace events to the frontend via SSE in real time. Firestore stores campaign history schemalessly to match our evolving lead data structure. The entire pipeline runs on Cloud Run with zero cold start issues at demo scale.

Next ProjectCORTEXOS →