D365 F&O MCP Server

GitHub Copilot with full knowledge of your D365 F&O codebase

What Is This?

An MCP server that gives GitHub Copilot access to your D365 F&O codebase — 584,799+ symbols indexed in SQLite FTS5. Copilot then knows exact method signatures, field types, class hierarchies, and generates X++ code that compiles on the first try.

┌─────────────────────┐    MCP/HTTP     ┌──────────────────────────────────┐
│  GitHub Copilot     │ ◄────────────► │  D365 F&O MCP Server             │
│  (Agent Mode)       │                 │                                  │
│                     │                 │  ┌─────────┐   ┌──────────────┐ │
│  "Show methods of   │                 │  │ Symbols │   │ Labels       │ │
│   SalesTable"       │                 │  │  ~1 GB  │   │  ~8 GB       │ │
│                     │                 │  │ 584K+   │   │ 19M+ labels  │ │
│  ← answer in 50ms   │                 │  │ symbols │   │ 70 languages │ │
└─────────────────────┘                 │  └─────────┘   └──────────────┘ │
                                        └──────────────────────────────────┘

| Without this server | With this server | |---------------------|-----------------| | Copilot guesses method names → compile errors | Exact signatures → code works | | Searching in AOT: 5–30 minutes | Answer in < 50 ms | | 584,799 symbols with no fast search | Everything indexed and instantly available |

Quick Start

git clone https://github.com/dynamics365ninja/d365fo-mcp-server.git
cd d365fo-mcp-server
npm install
copy .env.example .env          # Fill in PACKAGES_PATH and CUSTOM_MODELS
npm run extract-metadata         # Extract XML from D365FO packages
npm run build-database           # Build SQLite index (~5–20 min)
npm run dev                      # Server running at http://localhost:3000

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

GitHub Copilot Setup

1. Enable Editor Preview Features at github.com/settings/copilot/features

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

3. Add .mcp.json to the root of your Visual Studio solution (next to the .sln file):

{
  "servers": {
    "d365fo-code-intelligence": {
      "url": "http://localhost:3000/mcp/"
    },
    "context": {
      "workspacePath": "K:\\AosService\\PackagesLocalDirectory\\YourModel"
    }
  }
}

The server automatically locates your .rnrproj, extracts the correct model name, and writes new files to the right place in AOT. workspacePath is the only setting most users ever need.

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

4. Copy Copilot instructions to your solution root:

Copy-Item -Path ".github" -Destination "C:\Path\To\YourSolution\" -Recurse

What It Can Do — 29 Tools

🔍 Code Search & Navigation

| Prompt | What happens | |--------|-------------| | What methods does SalesTable have? | Returns all methods with their signatures | | Find classes related to invoicing | Full-text search across 584K+ symbols | | Where is CustTable.validateWrite() used? | Where-used analysis across the entire codebase | | Show me the structure of CustTable | Fields with EDTs, indexes, foreign key relations |

⚡ Code Generation

| Prompt | What happens | |--------|-------------| | Create a CoC extension for SalesTable.validateWrite() | Exact signature + boilerplate | | How is LedgerJournalEngine initialized? | Patterns from your own codebase | | Is my class MyHelper missing any standard methods? | Class completeness analysis | | Create a batch job for order processing | Complete batch job template |

🎨 Smart Object Generation

| Prompt | What happens | |--------|-------------| | Generate a transaction table with common fields | AI-driven table with intelligent field/index suggestions | | Create a SimpleList form for MyOrderTable | AI-driven form with datasource and grid controls | | What EDT should I use for CustomerAccount field? | Suggests EDTs using fuzzy matching and pattern analysis |

🏷️ Label Management

| Prompt | What happens | |--------|-------------| | Find a label for "customer" | Searches all AxLabelFile objects | | Translations of label MyFeature in model MyModel | All languages at once | | Create a new label MyNewField in MyModel | Writes to all .label.txt files |

📝 File Operations (local VM only)

• Generates correct D365FO XML for classes, tables, forms, enums
• Writes the file directly to the right location in AOT
• Automatically adds the file to .rnrproj
• Creates a backup before every change

Azure Deployment

Host the server on Azure App Service — the entire team shares a single instance.

| Resource | Configuration | Monthly cost | |----------|---------------|-------------| | App Service P0v3 | 1 vCPU, 4 GB RAM | ~$62 | | Blob Storage | 2 GB (symbols + labels) | ~$3 | | Redis Cache (optional) | Basic C0 | ~$16 | | Total without Redis | | ~$65 / month |

The database is automatically downloaded from Azure Blob Storage on server startup.

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

Documentation

| File | Contents | |------|---------| | docs/SETUP.md | Installation, configuration, Azure deployment | | docs/MCP_CONFIG.md | .mcp.json reference — workspace paths, UDE, project settings | | docs/MCP_TOOLS.md | All 29 tools with example prompts | | docs/USAGE_EXAMPLES.md | Practical examples: search, generation, CoC | | docs/ARCHITECTURE.md | Technical architecture, dual-database design | | docs/CUSTOM_EXTENSIONS.md | ISV / custom model configuration | | docs/PIPELINES.md | Automated metadata extraction via Azure DevOps |