Mendix Expert MCP Server

A self-learning, auto-researching AI assistant that gives your AI deep Mendix expertise and grows smarter with every interaction.

Get Started View on GitHub

npm

About this project

This is a Model Context Protocol (MCP) server that supercharges AI assistants (GitHub Copilot, Claude, ChatGPT, n8n, etc.) with deep Mendix expertise. As of v3.5.1, ALL clients participate in universal self-learning.

Key Capabilities

Capability	Description
Universal Self-Learning	ALL clients (Copilot, Claude, ChatGPT, n8n) get quality signals
REST /learn API	ChatGPT and automation tools can add knowledge via HTTP
Supabase Storage	242+ entries in PostgreSQL - survives Railway restarts
Semantic Search	253 vectors in Pinecone (OpenAI embeddings, 1536 dims)
Quality Assessment	Every search returns `answerQuality` and `beastModeNeeded`
Project Analysis	Analyzes your actual `.mpr` files (local MCP only)
Beast Mode	Exhaustive 5-tier research protocol for hard questions
Auto-Deploy	Push to GitHub → Railway deploys automatically

Quick Example

User: @mendix-expert How do I iterate over a list in a microflow?

AI: [Searches knowledge base]
    Answer Quality: strong | Web Search Recommended: No

    Based on the knowledge base, here are the patterns:
    1. Loop activity with IterableList...
    2. Aggregate with ListOperation...
    [Comprehensive answer with code examples]

Architecture Overview (v3.5.1)

┌─────────────────────────────────────────────────────────────────────┐
│                        AI CLIENTS                                   │
│  GitHub Copilot │ Claude Desktop │ ChatGPT │ n8n │ Make │ Zapier   │
└───────────────────────────┬─────────────────────────────────────────┘
                            │ MCP (stdio) or REST (HTTP)
                            ▼
┌─────────────────────────────────────────────────────────────────────┐
│                   Mendix Expert MCP Server v3.5.1                   │
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐                 │
│  │  Hybrid     │  │  Quality    │  │   Project   │                 │
│  │  Search     │  │ Assessment  │  │   Analyzer  │                 │
│  │             │  │  (v3.5.1)   │  │ (local only)│                 │
│  └─────────────┘  └─────────────┘  └─────────────┘                 │
│         │                │                                          │
│         ▼                ▼                                          │
│  ┌─────────────┐  ┌──────────────┐                                  │
│  │  Pinecone   │  │   Supabase   │◄─── Self-Learning                │
│  │ 253 vectors │  │  PostgreSQL  │     (all clients)                │
│  │  1536 dims  │  │ 242+ entries │                                  │
│  └─────────────┘  └──────────────┘                                  │
└─────────────────────────────────────────────────────────────────────┘
                            │
                            ▼ (auto-deploy)
┌─────────────────────────────────────────────────────────────────────┐
│  Railway Cloud: https://mendix-mcp-server-production.up.railway.app │
└─────────────────────────────────────────────────────────────────────┘

What’s New in v3.5.1

🌍 Universal Self-Learning - ALL clients get quality signals and can add knowledge
🧠 REST /learn endpoint - ChatGPT and automation can store discoveries
📊 Quality Assessment - Every search returns answerQuality and beastModeNeeded
🗄️ Supabase-first storage - 242+ entries in PostgreSQL
🔮 253 vectors in Pinecone - OpenAI text-embedding-3-small (1536 dims)
🚀 Auto-deploy - Push to GitHub → Railway deploys automatically

Last updated: December 12, 2025