Ask anything about D365
get answers from real code & metadata
ALM XPP MCP has indexed the entire Dynamics 365 standard — every class, table, method and label.
Ask in plain English. Get expert answers from real source code. No D365 knowledge required.
Stop losing hours to D365 guesswork
Every D365 team faces the same friction. ALM XPP MCP removes most of it.
"How does SalesFormLetter work?"
Stack Overflow. Docs. 2 hours of reading unfamiliar code.
Ask in plain English → get purpose, methods, tables & data flow. 3 seconds.
"Which roles can write to SalesTable?"
Open AOT. Trace manually. 45 minutes of security archaeology.
Full Role → Duty → Privilege → EntryPoint chain. Instant.
"How long will this FDD take to build?"
Senior dev reads the full spec. 1 hour estimation meeting.
AI effort estimate with real D365 matching patterns. 30 seconds.
One question. Complete intelligence.
Pick any D365 object. Get purpose, methods, tables and data flow — instantly from real source code, not documentation.
Not Just for Developers
ALM XPP MCP is designed for every role on a Dynamics 365 project. Ask questions in plain English -- no X++ knowledge required.
Code faster, refactor safer
Find any D365 object, generate X++ queries and unit tests with real metadata, detect N+1 queries and upgrade risks before they hit prod.
- Search 200K+ objects in natural language
- Generate X++ queries, data entities & unit test stubs
- Analyse PR impact & post review comments
- Detect performance issues & upgrade breaks
- Match runtime errors to source locations
Understand the system, speak its language
Ask questions in plain English and get business-friendly explanations of any D365 object, process or field -- without reading X++.
- Explain any table or form in business terms
- Map full business processes (O2C, P2P...)
- Understand security roles & access rights
- Find labels in any language instantly
- Ask "what does this field do?" in plain English
Estimate, track, ship with confidence
Get AI-powered effort estimates, sprint changelogs and requirement coverage reports -- straight from your Azure DevOps backlog.
- Estimate effort for any Work Item in hours
- Generate test scenarios from acceptance criteria
- Build release changelogs from merged PRs
- Check if a PR covers all requirements
- Gap/Fit analysis: standard vs custom dev
With vs Without ALM XPP MCP
The most impactful tools -- what your AI assistant can do vs doing it manually.
| Task | Without | With ALM XPP MCP |
|---|---|---|
| Search any D365 object | Open AOT, expand 200K+ nodes, inspect each one manually | search_d365_code -> instant semantic results with full metadata |
| Understand an object (non-technical) | Read XML, cross-reference fields and docs -- developer knowledge required | get_object_details -> full metadata, explained in plain English by your AI assistant |
| Trace a security chain | Navigate Role -> Duty -> Privilege across dozens of AOT nodes by hand | trace_security_chain -> full tree + Mermaid diagram in seconds |
| Review a Pull Request | Manually check X++ changes, BP rules, upgrade notes and CoC conflicts | ado_analyze_pr_impact -> BP violations + upgrade risk + auto-posted PR comment |
| Create a DevOps task & branch | Open DevOps, fill form, name task, create branch, link work item | ado_create_task -> FDD/RDD naming, branch created, item linked -- one command |
| Estimate effort for a Work Item | Senior dev reads the spec, manually estimates based on experience | ado_estimate_effort -> AI estimate (Analysis/Dev/Test/Deploy) with KB signals |
| Find any label in any language | Open label files, search across modules, switch language contexts manually | search_labels -> 24M+ labels across 74 languages, by text or label ID instantly |
Cloud & Local -- One Toolbox
90 tools in the cloud, 121 on your machine. Most are the same tools running in both places -- the local server adds 36 that write, compile and deploy. Split for a reason: your code and your compiler never leave your machine.
Indexes the standard D365 codebase (200K+ objects, 25M+ cross-references, 24M+ labels in 74 languages). Read, analyze, generate-as-text -- no access to your VM, files or compiler. Works in GitHub Copilot, M365 Copilot, Cursor & Claude.
Every cloud tool, plus 36 that only work here: needs your local D365 SDK + metadata to write AOT files, compile X++ (xppc), sync the DB, deploy and scaffold VS projects -- actions impossible from a remote multi-tenant cloud.
Why the split? A shared cloud cannot reach a customer's D365 virtual machine, filesystem or X++ compiler. Read/generate work happens in the cloud; anything that writes or builds runs locally -- so source never gets uploaded.
Calculate Your Savings
See the exact dollar value ALM XPP MCP delivers based on your role and usage.
Your Setup
Your Monthly Impact
Get Started in 3 Steps
Create an account, get your token, paste one config -- or connect directly to Microsoft 365 Copilot.
Create Your Account
Sign up free. Works with Google, Microsoft or email. No credit card required during beta.
Create AccountGet Your API Token
Dashboard -> "Request API Token" -> approved within 24h. Token is valid 30 days, auto-renewable.
Token approval required for security.
Connect Your Client
Pick your IDE below and paste one config block. Or connect via Microsoft 365 Copilot -- no config file needed.
Add to .vscode/mcp.json or %USERPROFILE%\.mcp.json (Visual Studio 2022 17.14+):
// .vscode/mcp.json — secrets are prompted, so this file is safe to commit { "inputs": [ { "id": "almxpp_api_key", "type": "promptString", "description": "ALM XPP MCP - API token", "password": true }, { "id": "devops_pat", "type": "promptString", "description": "Azure DevOps PAT", "password": true } ], "servers": { "almxppmcp": { "type": "http", "url": "https://api.almxpp.com/mcp", "headers": { "X-API-Key": "${input:almxpp_api_key}", "D365-Custom-Model-Path": "C:\\MyProject\\Metadata", "D365-Standard-Model-Path": "C:\\PackagesLocalDirectory", "DEVOPS_ORG_URL": "https://dev.azure.com/MyOrg", "DEVOPS_PROJECT": "My Project", "DEVOPS_PAT": "${input:devops_pat}" } } } }
Only X-API-Key is required — delete the lines you do not need. Live-environment (D365FO-*) and telemetry (AppInsights-*) headers are generated for you by the builder on your dashboard.
Go to your dashboard after getting your token and click Install in VS Code -- config is pre-filled.
Go to Dashboard// .cursor/mcp.json — Cursor cannot prompt, so keep this file out of Git { "mcpServers": { "almxppmcp": { "url": "https://api.almxpp.com/mcp", "headers": { "X-API-Key": "YOUR_TOKEN", "D365-Custom-Model-Path": "C:\\MyProject\\Metadata", "D365-Standard-Model-Path": "C:\\PackagesLocalDirectory", "DEVOPS_ORG_URL": "https://dev.azure.com/MyOrg", "DEVOPS_PROJECT": "My Project", "DEVOPS_PAT": "YOUR_DEVOPS_PAT" } } } }
// claude_desktop_config.json — stdio only, needs Node.js 18+ { "mcpServers": { "almxppmcp": { "command": "npx", "args": ["-y", "almxppmcp", "--api-key", "YOUR_TOKEN", "--model-path", "C:\\MyProject\\Metadata"], "env": { "DEVOPS_ORG_URL": "https://dev.azure.com/MyOrg", "DEVOPS_PROJECT": "My Project", "DEVOPS_PAT": "YOUR_DEVOPS_PAT" } } } }
- Open make.powerapps.com -> your Copilot
- Settings -> AI Plugins -> Add a plugin -> From URL
- Paste:
https://api.almxpp.com/ai-plugin.json - Connection: API Key -> enter your token
- Publish -> 90 D365 cloud tools available in M365 Copilot
- Download manifest: agent.json
- Teams Admin Center -> Integrated Apps -> Upload custom app
- Select
agent.json-> assign to users or groups - ALM XPP MCP appears in users' M365 Copilot panel
Frequently Asked Questions
https://api.almxpp.com/ai-plugin.json) or deploy it as a Declarative Agent from Teams Admin Center. Both options require a valid ALM XPP MCP API token. See the M365 tab above for step-by-step instructions.X-API-Key or Authorization: Bearer. Tokens are PBKDF2-SHA512 hashed at rest (100k iterations), sessions use cryptographic random tokens, and user data is AES-256-GCM encrypted. Tokens expire after 30 days and require admin approval for security.Your Data, Your Machine
The cloud server indexes only the public Microsoft D365 codebase. Your proprietary code reaches it solely when you upload a custom model or connect a repository — and stays in storage isolated to your account.