AgentAssist: AI Real Estate Lead Qualification
Multi-agent platform for automated lead qualification & listing intelligence
Top outcome
Automated lead onboarding in under 5 minutes
Stack
OpenAI API · MCP (Model Context Protocol) · Node.js / TypeScript · React / Next.js
Overview
A multi-agent AI platform that automates real estate lead qualification, MLS listing curation, and preference learning. The system uses six specialized agents - Qualification, MLS Ingestion, Matching & Ranking, Preference Modeling, Communication, and Agent Copilot - coordinated by a workflow orchestrator. External capabilities (MLS queries, messaging, CRM) are exposed via MCP servers for clean modularity. Leads are onboarded through a conversational wizard, receive daily curated listing feeds with explainable match rationale, and provide structured feedback that continuously improves recommendation quality. Built as a scalable foundation for future AI agents (negotiation, pricing, scheduling).
The Problem
Real estate agents spend significant time manually qualifying leads, searching MLS listings, and following up for feedback - resulting in missed opportunities and poor lead conversion.
The Solution
Multi-agent AI platform with specialized agents for lead qualification, MLS ingestion, preference modeling, and communication - coordinated by an orchestrator and connected via MCP servers.
Impact
- 1Automated lead onboarding in under 5 minutes
- 2Daily MLS-driven listing recommendations with interactive feedback loop
- 350–80% reduction in manual MLS searching
- 4Compounding match quality through structured preference learning
Key Decisions
- Adopted multi-agent + MCP architecture over monolithic LLM pipeline - higher modularity, better observability, cleaner separation of concerns
- Exposed MLS and messaging via MCP servers instead of hardcoding into prompts
- Structured "like/dislike + why" feedback over passive behavior inference for faster preference modeling
Lessons Learned
Explicit feedback beats implicit inference early on - structured "why" dramatically accelerates preference modeling
MLS integration is compliance-heavy, not just technical - data usage and display rules must be configurable
Explainability ("Why this listing") directly impacts user engagement and perceived relevance
Agent + tool architecture scales better than prompt-only systems
Screens
1 / 2