How I Built an AI Cold Calling and Email SaaS That Finds Its Own Leads [Claude Code + VAPI + Resend]

This tutorial demonstrates building a fully automated B2B outbound sales pipeline using AI-assisted code (Clot code). It covers crawling leads, storing contact data, automated AI voice calls, plain-text cold email sequences, and a live dashboard. The video explains project structure, knowledge base, orchestrator logic, queueing, test modes, and integrations needed to run a resilient, scalable lead automation system.

– Knowledge base: agent persona, core script with branching logic, objection handling, company info, and three plain-text email templates for sequenced outreach.
– Lead acquisition: Python-based crawler that mimics Google Maps searches, extracts business name, phone, address and website, supports CLI search args and deduplication before database insertion.
– Outreach modules: FAP voice-calling integration with dynamic variables for personalization and a Resend email layer using plain-text sequences and a test mode for safe verification.
– Orchestration & reliability: central status column state machine, cron-based orchestrator, Bull MQ-backed queues, rate limiting and exponential backoff for retries, plus webhook handlers to log outcomes to a live dashboard.

Quotes:

What if your sales pipeline can run itself? Not partially, not with manual intervention, but fully automated.

It’s literally doing what a human would do… except it does it in minutes instead of days.

That one second detail is the difference between sounding like a spam call and sounding like a real person.