InboxIntel

InboxIntel homepage showcasing AI-generated emails, high open rates, and a clean CTA to start or watch a product demo
User dashboard with live stats on campaigns, reply rates, recent activity feed, and a visual pipeline of outreach progress.
Campaign analytics with performance chart, daily email volume vs replies, and a timeline of recent outreach activity.

Inspiration

Every startup, indie hacker, and founder knows the pain of cold outreach. You spend hours crafting the perfect email, second-guessing every sentence — only to get ghosted. We wanted to change that.

InboxIntel is our answer: a cold outreach engine that doesn’t just help you send emails — it helps you win replies. We set out to build a tool that combines AI, automation, and campaign intelligence to remove the guesswork and make every message count.

And honestly? We wouldn’t have even tried building something this powerful in a weekend without Bolt, Supabase, and OpenAI — the dream stack for shipping fast and smart.

What it does

InboxIntel is a full-stack, AI-powered cold outreach platform that helps you:

🚀 Plan Campaigns: Define clear goals and audience targets.
✍️ Generate & Optimize Emails: Use OpenAI to improve tone, clarity, and personalization.
🔐 Authenticate Securely: Log in with Google via Supabase Auth.
📤 Send via Gmail API: Connect Gmail to send emails natively.
📬 Track Replies Automatically: Detect replies in threads using Gmail metadata and log them in real-time.
✅ Label Outcomes: Manually tag responses (positive, not interested, maybe later) and add notes.
🔁 Follow Up Intelligently: Generate follow-ups with OpenAI, in context of prior replies + user notes.
📊 See Campaign Insights: View stats like reply rate and performance of different email variants.

InboxIntel turns cold outreach into a data-driven, personalized, and intelligent process — in minutes, not hours.

How we built it

We built InboxIntel with:

🧩 Bolt for frontend and logic. Bolt is hands-down the fastest way to build full-featured apps. With its reactive programming model, live coding environment, and built-in state management, we moved at the speed of thought. It felt unfair.
🔐 Supabase Auth (Google Sign-In) for secure, seamless login. Auth setup took minutes and integrated perfectly with our RLS-based data access.
⚡ Supabase Edge Functions for all backend logic: OpenAI calls, Gmail polling, thread tracking, and webhook processing. These deploy fast, scale effortlessly, and keep our logic snappy and stateless.
✉️ Gmail API for sending emails and polling for replies. We store message/thread IDs in Supabase to link replies accurately.
🧠 OpenAI for drafting and optimizing emails — and generating smart follow-ups based on conversation history + user notes.

The whole app is reactive, real-time, and super responsive — thanks to this tightly integrated stack.

Challenges we ran into

Thread tracking via Gmail is no joke. Emails don't always play nicely with metadata, so we had to build fallback logic that matches by threadId and In-Reply-To.
OpenAI prompt tuning for follow-ups was nuanced — we wanted responses that felt genuinely human, not robotic. We got there with layered prompts and a touch of temperature tuning.
Edge Function latency and managing secure API keys in a hackathon setting required some orchestration, but Supabase made this 10x easier than other serverless platforms.

Accomplishments that we're proud of

A fully working cold outreach flow — from campaign setup to tracked follow-ups — all built and deployed in under 48 hours.
Seamless Gmail integration that feels native, not bolted on.
A genuinely useful product that we would use (and will).
A clean UX powered by Bolt — we moved fast without compromising quality.
Modular, scalable backend architecture using only Supabase Edge Functions and OpenAI.

What we learned

Bolt is our new go-to for building anything. We’ve used everything from Next.js to Flutter, but Bolt lets you think and build at the same time. We’ll be using it in production for future tools.
Supabase is way more than a Firebase clone — RLS + Auth + Edge Functions is a backend superpower combo.
Prompt design is product design. The right prompt is more powerful than any if-statement or state machine.
Speed ≠ rush — when your stack empowers you, you can move fast and still craft a high-quality experience.

What’s next for InboxIntel

We're just getting started. Next up:

🤖 AI-powered reply classification — automate tagging based on sentiment + intent.
🧠 Lead enrichment — pull job titles, locations, and funding status via Clearbit or LinkedIn scraping.
📅 Calendar integration — auto-schedule based on replies.
🔁 Autopilot follow-up sequences — triggered after X days without reply.
🧪 Prompt A/B testing — track conversion rates and optimize subject lines + variants.

We believe cold outreach can be radically smarter — and InboxIntel is just the beginning.