MR. AGENTMR. Agent — AI-Powered Sales Assistant with Interactive Forms and Knowledge Base

Why We Built MR. Agent

MR. Agent was born from a practical need: we wanted to provide 24/7 availability for potential clients while capturing high-quality leads through a more engaging experience than a standard contact form. Instead of building yet another chatbot, we leveraged our own open-source MDMA engine to create an AI agent that generates interactive forms, tables, and rich content dynamically during the conversation.

The system operates as an orchestrator — the LLM doesn't just answer questions, it calls server-side tools to search our knowledge base (portfolio projects, Clutch reviews, blog posts), retrieve company information, and even perform web searches restricted to our domain. Every response is grounded in verified data with strict anti-hallucination rules, ensuring the agent never fabricates project names, metrics, or case study details.

From Lead Capture to Sales Pipeline Integration

The primary goal was to create an AI-powered inbound sales channel that captures leads more effectively than static forms. The agent needed to support three distinct modes: knowledge Q&A about Mobile Reality (grounded in our portfolio, reviews, and blog content), guided project estimation with structured data collection, and quick contact for simple inquiries.

We aimed to demonstrate our own MDMA framework in production — proving that AI agents can dynamically generate interactive documents with forms, tables, charts, and approval workflows. MR. Agent serves as both a functional sales tool and a live showcase of what MDMA-powered solutions can deliver for our clients' businesses.

The system needed to integrate with our existing sales infrastructure — sending confirmation emails to users, notifying the sales team via Slack and email (in production), creating HubSpot CRM contacts automatically, and persisting complete conversation histories to DynamoDB for sales pipeline analysis.

Every interaction had to include appropriate disclaimers — the agent clearly identifies itself as AI, notes that pricing information is approximate, and directs users to consult with our team for final project estimates. This transparency builds trust while protecting against AI-generated inaccuracies.

AI Agents That Converse, Collect, and Convert

MR. Agent demonstrates how AI agents with interactive forms can transform the traditional lead capture experience. Instead of static form submissions, potential clients engage in a guided conversation where the AI adapts its responses based on context, generates dynamic forms for data collection, and grounds every answer in verified company knowledge.

The technical architecture — React frontend with dynamic MDMA rendering, NestJS orchestrator with tool-calling loop, Convex vector knowledge base, and DynamoDB session persistence — provides a production-grade blueprint for building AI sales assistants that integrate with existing CRM and communication infrastructure.

Built on our open-source MDMA framework, MR. Agent serves as a live demonstration of what conversational AI with structured interactive components can achieve. The same architecture powers our AI Headless Forms solution, enabling any business to replace rigid multi-step forms with adaptive chat experiences that converse, collect data, and convert.

The project validates a key thesis: AI-assisted development dramatically compresses delivery timelines. The MR. Agent module — including the MDMA playground component, backend orchestrator, knowledge base integration, email verification, and full sales pipeline integration — was built and iterated on rapidly using AI pair programming, demonstrating the development velocity that our AI-first approach delivers for client projects.