D365 F&O MCP Server

<div align="center">

54 AI tools that know every X++ class, table, method, and EDT in your D365FO codebase

Built for D365FO developers who write X++ in Visual Studio — not for generic web dev

</div>

---

Why this exists

AI coding assistants excel at C#, Python, and JavaScript — languages with rich public training data. X++ is different. Your D365FO codebase is private, highly customized, and deeply interconnected: thousands of CoC extensions layered over standard Microsoft code, ISV packages adding their own tables and classes, custom EDTs that drive field validation across dozens of forms. No AI has seen any of it.

The result is AI that confidently generates code that doesn't compile: wrong method signatures, missing parameters, fields that don't exist on the table, CoC chains broken because the AI didn't know an extension already wrapped the method.

This MCP server solves that by pre-indexing your entire D365FO installation — hundreds of thousands of symbols across standard and custom models — and exposing it as 54 specialized tools. Works with GitHub Copilot and Claude Code CLI. Before generating any X++ code, the AI can look up exact method signatures, check what CoC extensions already exist, trace security hierarchies, find label translations, and understand the full shape of your data model. The result is code that compiles on the first try and integrates correctly with your existing customizations.

Without this server	With this server
AI guesses method signatures → compile errors	Exact signatures pulled from your actual codebase
"Does CustTable.validateWrite() have any CoC wrappers?" requires manual AOT search	find_coc_extensions answers in < 50 ms
ISV and custom model extensions invisible to AI	All models fully indexed and searchable
Security hierarchy takes hours to trace manually	get_security_coverage_for_object traces Role → Duty → Privilege → Entry Point instantly
AI generates hardcoded strings instead of label references	search_labels finds the right @SYS/@MODULE label key immediately
"Which tables reference CustTable?" requires digging through AOT relations	get_table_relations returns every FK relation and cardinality in one call
EDT base types and field lengths are a constant lookup	get_edt_details returns the full EDT definition including extends chain
AI doesn't know which SysOperation framework class to extend	search_classes with a description filter surfaces the right base class
New CoC extension may silently duplicate an existing one	find_coc_extensions reveals all existing wrappers before you write a line
Menu item to form mapping requires navigating the AOT manually	get_menu_item_details resolves the full path from menu item to form to data source

---

Key Capabilities

• Massive Metadata Index: Instantly looks up signatures, tables, enums, EDTs across standard and ISV code out of hundreds of thousands of objects.
• Smart Object Generation: AI-driven tools build XML structures exactly matching D365 standard patterns (SimpleList, Forms, Tables).
• X++ Knowledge Base: Queryable knowledge base of D365FO patterns, best practices, and AX2012→D365FO migration guidance — prevents deprecated API usage.
• Code Review & Git Diff: Reviews uncommitted workspace changes locally matching D365 Best Practice metrics directly against Copilot chat.
• D365 SDLC Native Integration: Triggers local MSBuild, sync.exe database updates, SysTestRunner tests, and xppbp.exe best practice validations smoothly behind the scenes.
• XML Parsing without Corruption: Intelligently mutates AxTable and AxForm files locally avoiding the common string replacement corruption associated with VS agents.

Quick Start

Full step-by-step guide with all scenarios: docs/QUICK_START.md

Prerequisites — install required software via d365fo.tools:

Install-D365SupportingSoftware -Name vscode,python,node.js

git clone https://github.com/dynamics365ninja/d365fo-mcp-server.git
cd d365fo-mcp-server
npm install

# Build the C# bridge (required on Windows D365FO VMs for file create/modify)
cd bridge\D365MetadataBridge
dotnet build -c Release
cd ..\..                         # Back to repo root

copy .env.example .env           # Set PACKAGES_PATH, CUSTOM_MODELS, LABEL_LANGUAGES, ...
npm run extract-metadata         # Extract XML from D365FO packages
npm run build-database           # Build SQLite index
npm run dev

UDE / Power Platform Tools? Run npm run select-config instead of setting PACKAGES_PATH manually.

---

Connect to GitHub Copilot

1. Enable MCP servers in Copilot at github.com/settings/copilot/features

2. In Visual Studio: Tools → Options → GitHub → Copilot → check Enable MCP server integration in agent mode

3. Create %USERPROFILE%\.mcp.json (covers all solutions on the machine, recommended) or place .mcp.json next to a specific .sln file:

{
  "servers": {
    "d365fo-azure": {
      "url": "https://your-server.azurewebsites.net/mcp/"
    },
    "d365fo-local": {
      "command": "node",
      "args": ["K:\\d365fo-mcp-server\\dist\\index.js"],
      "env": {
        "MCP_SERVER_MODE": "write-only",
        "D365FO_SOLUTIONS_PATH": "K:\\VSProjects\\MySolution",
        "D365FO_WORKSPACE_PATH": "K:\\AosService\\PackagesLocalDirectory\\YourPackageName\\YourModelName"
      }
    }
  }
}

4. Copy the Copilot instruction files so Copilot knows the D365FO workflow rules:

# Place .github in a parent folder shared by all your D365FO solutions, e.g.:
Copy-Item -Path ".github" -Destination "C:\source\repos\" -Recurse

Tip: Visual Studio 2022 searches for .github\copilot-instructions.md upward from the solution folder, so one copy in a common parent directory covers all solutions underneath — no need to copy it into every solution separately.

Full config options (UDE paths, explicit projectPath, solutionPath): docs/MCP_CONFIG.md

---

Connect to Claude Code CLI

Claude Code uses a different config format from Copilot — "mcpServers" key, explicit "type" field, and "alwaysLoad": true to prevent tool deferral.

1. Install Claude Code CLI: npm install -g @anthropic-ai/claude-code

2. Register the server (writes to ~/.claude.json):

# Azure-hosted
claude mcp add-json --scope user d365fo-mcp-tools '{"type":"http","url":"https://your-server.azurewebsites.net/mcp/","alwaysLoad":true}'

# Local stdio
claude mcp add-json --scope user d365fo-mcp-tools '{"type":"stdio","command":"node","args":["K:\\d365fo-mcp-server\\dist\\index.js"],"env":{"DB_PATH":"K:\\d365fo-mcp-server\\data\\xpp-metadata.db","LABELS_DB_PATH":"K:\\d365fo-mcp-server\\data\\xpp-metadata-labels.db","D365FO_SOLUTIONS_PATH":"K:\\repos\\MySolution\\projects","D365FO_WORKSPACE_PATH":"K:\\AosService\\PackagesLocalDirectory\\YourPackageName\\YourModelName"},"alwaysLoad":true}'

alwaysLoad: true loads the d365fo tool list at session start, preventing Claude from routing X++ lookups to other connected code intelligence tools or built-in search.

3. Copy CLAUDE.md to the parent folder of your D365FO solutions:

Copy-Item -Path "K:\d365fo-mcp-server\CLAUDE.md" -Destination "C:\source\repos\"

Full Claude Code setup guide (all scenarios, project-scoped .mcp.json, troubleshooting): docs/CLAUDE_CODE_SETUP.md

---

Azure Deployment

Host on Azure App Service so the whole team shares one instance — nobody needs the server running locally.

Resource	Configuration	Monthly cost
App Service Basic B3	4 vCPU, 7 GB RAM	~$52
Blob Storage	~2.5–3.5 GB (symbols + labels, without/with UnitTest models)	~$3
Total		~$55 / month

The database downloads from Azure Blob Storage automatically on startup.

Setup guide: docs/SETUP.md · CI/CD pipeline: docs/PIPELINES.md

---

Documentation

File	Contents
docs/QUICK_START.md	Start here — 5 steps to get running, all .mcp.json parameters, logging
docs/SETUP.md	Detailed installation, configuration, all deployment scenarios A–F
docs/MCP_CONFIG.md	.mcp.json reference — workspace paths, UDE, project settings, all env vars
docs/MCP_TOOLS.md	All 54 tools with parameters and example prompts
docs/USAGE_EXAMPLES.md	Practical examples: search, CoC, SysOperation, security
docs/CUSTOM_EXTENSIONS.md	ISV / custom model configuration and multi-model extraction
docs/WORKSPACE_DETECTION.md	How the server auto-detects your D365FO project, model, and package path
docs/ARCHITECTURE.md	Technical architecture, dual-database design, cache invalidation
docs/BRIDGE.md	C# Metadata Bridge reference — mandatory on Windows VMs for all write operations
docs/SETUP_AZURE.md	Deploy the server to Azure App Service (admin/DevOps guide)
docs/PIPELINES.md	Automated metadata extraction and deployment via Azure DevOps
docs/TESTING.md	Running tests, test structure, mock guidelines, coverage
docs/SQLITE_DEPENDENCY.md	SQLite vs C# Bridge — which tools use which data source
docs/CLAUDE_CODE_SETUP.md	Connecting Claude Code CLI — .mcp.json + CLAUDE.md setup