CensusChat
hollandkevint/CensusChat
Consultas demográficas do Censo via linguagem natural para healthcare
Ver no GitHubInterface em linguagem natural para dados do Censo dos EUA, construída com Claude AI, DuckDB e MCP. Responde perguntas sobre demografia e saúde em segundos, sem necessidade de conhecimento em SQL. Inclui tabelas interativas, gráficos e exportação em Excel/PDF para equipes de estratégia healthcare.
Instalação
npx convez-ai hollandkevint/CensusChat1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Análise de elegibilidade Medicare por região geográfica
- Comparação paralela de demográficas entre mercados (ex: Tampa vs Phoenix)
- Estratégia de expansão e posicionamento de unidades de saúde
- Pesquisa em saúde populacional com drill-down de county a block group
Veredito

Solução sólida de acesso a dados estruturados via IA, com implementação real de MCP e segurança (SQL injection prevention). Bem executado para o segmento healthcare, mas com mercado inicial bastante nicho (equipes de estratégia de sistemas de saúde e Medicare Advantage).
README
A tradução para o português deste README ainda está na fila.
CensusChat: Healthcare Demographics Made Simple
Natural language interface to US Census data for healthcare strategy teams
Transform 6-week $50K demographic consulting into 6-second $297/month queries
📊 Foundation Data • 📖 Quick Start • 📧 Get Early Access
🎉 v1 Shipped (February 3, 2026)
What's New in v1
- DuckDB 1.4 — New async API, MERGE statements, encryption support, query profiler
- MCP HTTP Transport — External client connections (Claude Desktop, Postman)
- Interactive Data Tables — TanStack Table with sorting, filtering, pagination, drill-down
- Chart Visualizations — Recharts bar/line charts with export controls
- Claude Agent SDK — Zod schema validation, parallel region comparison, conversational context
- Document Generation — Excel and PDF reports via MCP tools
Core Features
- ✅ TRUE MCP IMPLEMENTATION: Production SQL validation layer with security policies
- ✅ SQL Injection Protection: Table/column allowlists, row limits, dangerous pattern blocking
- ✅ Interactive Results: Sortable tables, filterable data, drill-down navigation
- ✅ Parallel Queries: "Compare Tampa Bay vs Phoenix" runs both simultaneously
- ✅ Conversational Context: "Now filter to income > $75K" understands prior query
- ✅ Audit Compliance: All queries logged for HIPAA/GDPR
Current Data:
- County Level: 3,144 US counties with demographics
- Block Group Level: 239,741 block groups with 84 variables
🎯 The Problem
Healthcare strategy teams wait 6+ weeks and pay $50K+ for demographic analysis that CensusChat delivers in seconds for $297/month.
"Compare Medicare eligible seniors in Tampa Bay vs Phoenix with income over $75K"
→ Parallel queries, interactive comparison table, instant Excel export
⚡ How It Works
1. Natural Language Input
Ask questions in plain English - no SQL knowledge required
2. AI-Powered Processing with Security
- Claude Sonnet 4 translates natural language to SQL
- MCP Validation Layer validates SQL against security policies
- DuckDB 1.4 executes validated queries on Census data
- Agent SDK handles parallel queries and conversational context
3. Interactive Output
- Sortable, filterable data tables
- Bar and line chart visualizations
- Drill-down from county to block groups
- Excel/PDF export with one click
🏥 Built for Healthcare Professionals
Primary Users:
- Business Analysts at health systems and Medicare Advantage plans
- Strategy Teams planning market expansion and facility placement
- Healthcare Researchers conducting population health studies
🛠 Technical Stack
const stack = {
backend: "Node.js 20 + TypeScript + Express 5",
frontend: "Next.js 15 + React 19 + Tailwind CSS 4",
databases: "PostgreSQL 15 + Redis 7 + DuckDB 1.4.3",
ai: "Claude Sonnet 4 + Agent SDK 0.2.30 + MCP SDK 1.25.3",
interactive: "TanStack Table + Recharts + MCP Apps",
deployment: "Docker + containerized testing"
};
Performance Metrics
- Sub-2 Second Queries with MCP validation
- Parallel Execution for comparison queries
- 82%+ Test Coverage with comprehensive test suites
- HIPAA-Ready Architecture with encryption and audit logging
🚦 Quick Start
Step 1: Get API Keys (Required)
See API_KEY_SETUP.md for detailed instructions.
- Anthropic API: Get key at console.anthropic.com
- Census API: Request key at api.census.gov/data/key_signup.html
- Add both keys to
backend/.env
Step 2: Load Census Data (First Time Only)
cd backend
./scripts/setup-database.sh # Takes 2-3 hours
Step 3: Start the Application
./demo-setup.sh
Open http://localhost:3000 and start querying.
📊 Features
Healthcare-Specific Queries
- Medicare eligibility calculations (age 65+ demographics)
- Senior care demographics for facility planning
- Population health indicators and social determinants
- Market analysis for health system expansion
Interactive Data Exploration
- Click column headers to sort
- Filter rows without re-querying
- Drill down from county to block groups
- Export selected columns to Excel/CSV
Comparison Queries
- "Compare Tampa Bay vs Phoenix" runs in parallel
- Side-by-side demographic breakdowns
- Automatic chart generation for comparisons
Conversational Follow-up
- "Show seniors in Tampa Bay"
- "Now filter to income over $75K" (understands context)
- "Export that to Excel"
Enterprise Security
- HIPAA-ready architecture with encryption at rest
- SQL injection protection with validation layer
- Complete audit trail for compliance
- No PHI storage - only aggregated demographic data
📚 Documentation
Getting Started
Technical Guides
Deployment
📈 Market Impact
Problem We Solve:
- $2.8B spent annually on demographic consulting
- 6,090 hospitals needing demographic analysis
- 30,000+ senior care facilities planning expansion
Our Solution:
- 200x cost reduction: $50K → $297/month
- 300x speed improvement: 6 weeks → seconds
- Unlimited iterations with conversational follow-up
🚀 Get Involved
For Healthcare Teams
For Developers
- Write quality TypeScript with 90%+ test coverage
- Maintain sub-2s query response times
- Follow HIPAA-ready security practices
- Run
npm run secret-scanbefore committing
⭐ Star this repository if CensusChat helps your healthcare data work
Transforming healthcare demographic analysis from weeks to seconds
👤 About Kevin Holland
10+ years healthcare data experience • Frustrated by $50K consulting delays