fo-semantic-mcp

Overview

What is fo-semantic-mcp?

fo-semantic-mcp is a RAG (Retrieval-Augmented Generation) knowledge server for D365 F&O AI development. It connects your AI coding assistant to an enriched F&O artifact knowledge base — not raw AOT exports, but a semantically indexed, AI-enhanced representation of 150,000+ artifacts with 19M+ cross-reference relationships extracted from the F&O compiler.

Built on the Model Context Protocol (MCP), it gives your AI grounded F&O knowledge it can't get from training data alone. When you're developing X++ extensions or customizations, your AI IDE (Cursor, Claude Code, VS Code) automatically retrieves the exact artifacts, patterns, and dependencies it needs — reducing hallucination and enabling accurate code generation.

This is what makes AI pair programming real: your AI doesn't guess how Microsoft built SalesTable — it retrieves the actual implementation, cross-references, and business context from the index.

⚠️ Important: This is a developer tool for accessing F&O source code, metadata, and artifacts (classes, tables, forms, etc.). It is not designed for accessing F&O business data or database records. Use this MCP server to understand code structure, learn implementation patterns, and develop extensions — not to query customer orders, inventory data, or other business information.

Knowledge

Beyond Raw AOT — What Your AI Actually Gets

The artifact index is not a dump of AOT XML files. It's a purpose-built, AI-enriched knowledge base that gives your coding agent the context it needs to write correct X++ code:

Want to explore the index without an IDE? Use Artifact Search — the same database, with a browser interface for quick lookups and demos.

Why Semantic Search

Why This Beats Keyword Search for X++ Development

Integration

Getting Started

1. Get Your API Key

   Sign in to xplusplus.ai and generate your API key from the dashboard.

2. Add Configuration to Your IDE

   Add the MCP server configuration - npx will automatically download and run it:

   Published on the official MCP registry.

3. Configure Your IDE

   Add the MCP server to your IDE configuration file:

   Cursor IDE:

   Edit ~/.cursor/mcp.json

   ~/.cursor/mcp.json

   {
     "mcpServers": {
       "fo-semantic-mcp": {
         "command": "npx",
         "args": ["-y", "fo-semantic-mcp"],
         "env": {
           "FOINDEX_API_KEY": "your_api_key_here",
           "FO_LOCAL_ASSETS_PATH": "C:\\Users\\YourName\\AppData\\Local\\Microsoft\\Dynamics365\\10.0.xxxx\\PackagesLocalDirectory"
         }
       }
     }
   }

   Tip: Add FO_LOCAL_ASSETS_PATH to enable reading actual F&O XML source files

   No installation needed! npx automatically downloads and runs the latest version.

   Other AI Tools Setup Guides

   Claude Code

   CLI & VS Code Extension

   Setup Guide →

   Claude Desktop

   Standalone Desktop App

   Setup Guide →

   OpenAI Codex CLI

   Terminal Agent

   Setup Guide →

   Google Antigravity

   Agentic VS Code IDE

   Setup Guide →

   Microsoft Copilot Studio

   Low-Code AI Agent Builder

   Setup Guide →
4. Restart Your IDE

   Restart Cursor IDE or Claude Desktop. npx will automatically download the MCP server on first run (takes a few seconds), then it's cached for instant startup.

5. Start Coding with AI

   Your AI assistant can now automatically access F&O artifacts, find patterns, and suggest implementations using natural language.

Reference

Configuration Options

C:\Users\[YourName]\AppData\Local\Microsoft\Dynamics365\[version]\PackagesLocalDirectory

Get-ChildItem "$env:LOCALAPPDATA\Microsoft\Dynamics365\" -Recurse -Filter "PackagesLocalDirectory" | Select-Object FullName

Workflow

How It Works in Practice

🎯 Automatic AI Pair Programming (Most Important)

When you're writing X++ code for extensions or customizations, your AI IDE automatically calls the MCP server to retrieve out-of-box F&O objects. You don't need to manually search — your AI assistant proactively:

• Fetches the original class/table/form you're extending
• Analyzes existing patterns and methods
• Suggests proper extension approaches based on actual F&O code
• References related artifacts and dependencies automatically

Example: When extending SalesTable, Cursor IDE automatically retrieves the SalesTable class definition, studies its methods, and suggests how to properly implement your extension.

📝 Manual Queries

You can also explicitly ask your AI assistant questions like:

"Find customer payment processing examples"
"Show sales order validation patterns"
"How do I implement inventory transactions?"
"Read the SalesTable form XML"

The MCP server translates these queries into precise artifact searches and delivers the results directly to your AI coding assistant.