Leadchat.ai

Small B2B teams in the UK, EU and US lose inbound leads at night, at weekends and whenever a manager is busy, and spend hours asking the same qualifying questions. Leadchat.ai is my own product built for them: an AI agent that answers on every messaging channel within seconds, speaks in the company’s voice and hands over only warm leads. It also had to be cheap to operate with a small team.

I designed Leadchat.ai as an event-driven system: a Laravel 11 + Inertia/Vue 3 core for accounts, leads and billing, plus eight TypeScript services on Bun for AI orchestration, channel webhooks and senders, CRM sync, a realtime WebSocket gateway, speech-to-text and meeting billing, all connected through Redis Streams consumer groups. Agents follow a configurable qualification script, book Zoom or Google meetings and sync results to HubSpot or Pipedrive. Production runs on AWS EKS provisioned with Terraform, with GitHub Actions CI and GlitchTip monitoring.

Channels: WhatsApp, Messenger, Instagram, Telegram Business, web widget

Guided onboarding: company, website widget, agent review, CRM and calendar

Ready-made industry agent personas or custom agents with knowledge and tone

CRM sync with HubSpot, Pipedrive and OneBox; Zoom and Google meeting booking

OpenAI chat and Realtime voice, Deepgram STT, HeyGen avatars for demo calls

Live at leadchat.ai with 20+ integrations and multi-tenant, multi-currency usage-based billing

Laravel 11

Inertia + Vue 3

TypeScript / Bun

PostgreSQL

Redis Streams

OpenAI

AWS EKS

Terraform

GitHub Actions