AgentFlow - Jeremy Dawes

Overview

A complete voice AI agent platform that combines conversational AI with full CRM capabilities. Built for businesses deploying ElevenLabs voice agents who need call tracking, transcription, lead capture, and customer management in one system.

Database Tables

CRM Modules

Voice Features

50+

API Endpoints

The Problem

Deploying voice AI agents creates operational challenges:

Call data scattered across multiple systems
No connection between voice interactions and CRM records
Manual transcription and analysis of conversations
Lead capture from calls requires human intervention
No unified view of customer journey

Businesses need their AI agents integrated with their business systems.

Solution

Full-stack platform connecting voice AI to business operations:

Voice AI4

Calls, transcripts, analysis, webhooks

CRM6

Contacts, companies, deals, cases

Content3

Notes, docs, attachments

Platform4

Auth, API tokens, settings

Voice AI Integration

ElevenLabs Voice Agents

Feature	Description
Call Logging	Every conversation captured with metadata
Transcription	Full transcript with speaker attribution
AI Analysis	Sentiment, topics, action items extraction
Webhook Processing	Real-time event handling

Call Data Captured

Field	Description
Duration	Call length in seconds
Cost	ElevenLabs API cost tracking
Status	initiated, in_progress, completed, failed
Termination Reason	Why the call ended
Recording	R2 storage path
Transcript	Searchable message history

AI Analysis Results

Analysis	Description
Summary	AI-generated call summary
Sentiment	positive, neutral, negative, mixed
Topics	Extracted discussion topics
Action Items	Follow-up tasks identified
Data Extraction	Name, email, phone, company from transcript

CRM Modules

Contacts & Companies

Full people and organisation management with:

Social links (LinkedIn, Facebook, Instagram)
Address management with Australian postcode support
Tags and categorisation
Status tracking (active, inactive, archived)

Deals Pipeline

Kanban-style sales pipeline with:

Drag-and-drop stage management
Deal values and probabilities
Expected close dates
Contact and company linking

Cases

Support ticket system with:

Priority levels (low, medium, high, urgent)
Type categorisation (support, sales, billing, technical)
Due dates and assignment
Entity linking to contacts, companies, deals

Notes System

Rich text notes with:

TipTap editor with formatting
Threaded replies (single-level)
File attachments via R2
Polymorphic linking to any entity

Lead Capture

Public Enquiry Forms

Simple and detailed form variants
Cloudflare Turnstile captcha protection
Automatic IP and user agent logging
Status workflow (new → contacted → qualified → converted)

Enquiry to CRM Conversion

One-click conversion creates:

New contact from enquiry data
New company if provided
Optional deal creation
Full audit trail maintained

Architecture

AgentFlow Platform
    ↓
React 19 SPA + Hono API
    ↓
Cloudflare Workers + D1
    ↓
Integrations:
├── ElevenLabs (voice agents)
├── R2 (file storage)
├── better-auth (authentication)
└── TipTap (rich text)

Tech Stack

Frontend: React 19, Vite 7, Tailwind v4, shadcn/ui
Backend: Hono on Cloudflare Workers
Database: D1 (SQLite at edge)
ORM: Drizzle
Auth: better-auth with D1 adapter
Voice: ElevenLabs Conversational AI
Storage: Cloudflare R2
Rich Text: TipTap editor

Multi-Tenancy

Row-level security with userId isolation:

Every user-scoped table has userId foreign key
API middleware enforces ownership verification
Cascade delete on user removal
No cross-tenant data leakage

Use Cases

Voice Agent Operators - Track all calls, transcripts, and outcomes

Sales Teams - Connect voice interactions to deal pipeline

Support Desks - Link calls to cases and customer records

Lead Generation - Capture and qualify leads from AI conversations

Business Analytics - AI-analysed sentiment and topic trends