Several months ago, I started the Vibe Coding for AL initiative with Microsoft— publishing AL rules to help AI assistants “vibe” code with high-quality AL code that follows best practices.
Then something amazing happened — MVPs began experimenting, Microsoft engineers jumped in, and suddenly we weren’t just talking about coding rules anymore. We were building agentic development tools. Out of that collaboration, a whole ecosystem of MCP servers was born.
MCPs (Model Context Protocol servers) are the bridge between copilots and real developer tools. Instead of guessing (aka hallucinating), copilots can now query symbols, manage IDs, fetch translations, or even run telemetry queries — all thanks to MCPs.
And honestly — it’s changing the game for AL developers.
Let me take you through the MCPs that grew out of this journey (and a few 3rd-party ones).
🧑💻 BC Code Intelligence MCP
This one comes from Jeremy Vyska. Instead of a single generic copilot, you get 14 domain experts in Business Central — each with their own personality and specialty. Ask a coding question, and you’ll meet Sam Coder. Struggling with troubleshooting? Dean Debug is on it. Architecture decisions? Alex Architect steps in.
It feels less like one assistant and more like a team of specialists at your service.
Highlights:
14 personas (Sam Coder, Dean Debug, Alex Architect, Riley Report, etc.)
20+ tools for knowledge search, code reviews, workflow orchestration
Smart routing: queries automatically go to the right expert
Watch the episode of Dynamics Corner where Kris, Brad and Jeremy discusses the intersection of AI and Business Central development.
🔍 AL Dependency MCP Server
Stefan Maron built this to tackle one of the hardest things in AL: navigating dependencies. Ever tried to make sense of the massive 50MB+ Base App symbols? Now copilots can do that for you — instantly.
The server exposes compiled symbols to your copilot, so instead of guessing, it actually knows what objects exist, what fields are inside, and how they relate.
Highlights:
Handles huge symbol files with sub-100ms lookups
6 tools for object search, definitions, summaries, members, stats
Auto-discovers
.alpackagesin your workspace
🆔 AL Object ID Ninja MCP
If you’ve ever fought with object ID collisions, you’ll love this. If you are professional AL developer I guess AL Object ID Ninja, created back in 2021 by Vjeko, is in your toolbox.
Thanks, to Torben Leth Ninja has now leveled up into a full MCP server — part of the agent’s toolbox as well. It runs in two modes: Lite for individual devs, and Standard for teams. With Standard, you get backend sync, shared ID pools, and automatic conflict prevention across your whole team.
Highlights:
Lite mode (4 tools) for solo developers
Standard mode (8 tools) for teams with shared ID pools
Tools for ID allocation, workspace analysis, conflict detection
🌐 NAB AL Tools MCP
Translation is always tricky, and the Johannes Wikman from NAB Solutions took he’s NAB AL Tools and turned into an MCP server. That means copilots can now refresh XLIFF files, fetch untranslated texts, save translations, and even pull glossary terms — all through MCP tools.
Highlights:
8 tools for translation workflows
Refresh and sync XLF files
List untranslated texts or save new ones
Glossary term lookups for consistency
🆕 📊 BC Telemetry Buddy MCP
I’m super happy that our Troubleshooting Buddy, presented at BC TechDays, has found new life! Waldo created BC Telemetry Buddy MCP, which connects Copilot to your Application Insights data.
Instead of writing KQL queries by hand, you can just ask in natural language:
“Show me all login errors from the last 24 hours.”
The MCP backend translates that into the right query, pulls the data, and gives you results. It even includes a setup wizard and a built-in query library so you can organize and reuse your favorite queries.
Highlights:
10+ tools for querying telemetry with KQL
Event catalog & schema discovery
Save and organize queries by company
Natural language queries directly in VS Code
🆕 Native AL Extension Copilot Tools
As of September 30, 2025, the AL VSIX extension for VS Code now ships with a set of GitHub Copilot agent tools (pre-release). These aren’t MCP servers (yet) — they’re tightly coupled to GitHub Copilot and currently only work in VS Code (I couldn’t get them running in Cursor).
But still — it’s a big step forward. With these tools, Copilot Agent Mode can automate and orchestrate AL workflows directly in your editor:
al_build→ build your projectal_publish/al_fulldependencypublish→ publish appsal_downloadsource→ download source codeal_debug→ start debugging sessionsal_navigate→ navigate symbols…and more
👉 Try it here: AL Extension on Marketplace
⚡ 3rd Party MCPs
The AL-focused MCPs are game-changers, but there’s a growing set of general-purpose MCPs that blend beautifully into AL dev workflow.
💻 Serena MCP (with AL support!)
Serena is a multi-language MCP server that integrates with the Language Server Protocol (LSP). Thanks to Torben Leth, it now supports AL Language Server too. That means copilots can navigate your AL codebase with findReferences, goToDefinition, documentSymbols, and more.
📚 Microsoft Learn MCP
Launched earlier this year, the Microsoft Learn MCP connects copilots directly to official Microsoft Docs. It’s small and elegant — just 3 tools (search, fetch, code_sample_search) — it guarantees copilots won’t hallucinate when you need the official documentation.
🐙 GitHub MCP
This is the official MCP server from GitHub, giving copilots and agents direct access to your repos, issues, PRs, workflows, and code insights. Agent can now read, search, and manage your GitHub projects through natural language.
🚀 Azure DevOps MCP
This one comes straight from Microsoft. It exposes 50+ tools across work items, repos, pipelines, wikis, and search. If you’re serious about DevOps automation with Copilot, this is the official way to do it.
🕒 Clockify MCP
One more from Jeremy Vyska. Connects Copilot to Clockify. With 33 tools for managing projects, tasks, clients, and timers, you can track time and handle invoicing right from your AI assistant.
🔮 What’s Next: CentralQ MCP
Yes, it’s coming.
But unlike the other MCPs, the real challenge here isn’t the server itself — it’s the knowledge it returns.
Over the last 2.5 years of running CentralQ.ai, I’ve learned that raw indexing of blogs, videos, docs etc. isn’t enough. Agents don’t just need context — they need clean, structured, connected, community knowledge.
That’s why I’m rebuilding the backend from scratch:
Smarter pipelines for parsing community content
Re-indexing with richer metadata and LLM-based revisions
Delivering reliable high-quality context that agents and copilots can actually use
When it’s ready, CentralQ MCP will become the ultimate knowledge source of truth for every AL agent and copilot.
Stay tuned. 🚀
💡 Final thought
Vibe Coding started with rules, MCPs give us the tools. Together they’re shaping the future of AL development. And with CentralQ MCP around the corner — we’re only just getting started.
🔄 Update
Jeremy just published a fantastic new resource on Agentic Coding:
👉 Agentic Coding on alguidelines.dev
It explains how to work effectively with AI-powered assistants (agents) in AL development for Business Central.
