Microsoft Brings MCP to Azure Logic Apps (Standard) in Public Preview, Turning Connectors into Agent Tools
Microsoft Brings MCP to Azure Logic Apps (Standard) in Public Preview, Turning Connectors into Agent ToolsMicrosoft's latest public preview lets Azure Logic Apps Standard work as Model Context Protocol servers, making your existing workflows discoverable and callable by AI agents like VS Code Copilot. With over 1,400 connectors now potentially available as AI tools, enterprises can seamlessly bridge their automated processes with intelligent agents through secure, OAuth-protected endpoints.
Anandesh Sharma
·
Sep 24, 2025
Microsoft just dropped a pretty cool public preview feature - they've made it so Azure Logic Apps (Standard) can work as Model Context Protocol (MCP) servers. What this basically means is that your Logic Apps workflows can now show up as tools that AI agents can find and use automatically.
So if you're working in something like VS Code with Copilot, those AI assistants can now discover your Logic Apps workflows and actually run them when they need to. It's like giving AI agents direct access to your automated processes, which could be a game-changer for how we integrate AI into our existing workflows.
What’s actually shipping
Turn Your Logic Apps into AI-Powered ToolsThink of this as giving your Standard Logic Apps a special superpower. You set up an MCP endpoint (just a fancy URL at /api/mcp) and suddenly your HTTP workflows become smart tools that AI can use on demand. The best part? Microsoft handles all the security headaches with Easy Auth and OAuth 2.0 - it's like having a bouncer that automatically knows who's allowed in. Plus, if you're using VS Code 1.102 or newer, you can test everything right from your editor without breaking a sweat.
The Enterprise-Grade Approach via API CenterWant something more organized? Microsoft's got you covered with Azure API Center (still in preview). This is where things get really interesting - you can register your MCP servers here and watch as your managed connector actions transform into discoverable AI tools. It's like having a well-organized toolbox where everything is labeled, catalogued, and governed properly. Perfect for teams that need to keep their AI integrations neat and compliant.
Both paths give AI agents direct access to your existing workflows - it's like having a really smart assistant that knows exactly which tools to grab from your digital workshop.
Key requirements and transport details
Workflow shape: Tools must be implemented as HTTP Request trigger (“When a HTTP request is received”) plus a Response action.
Auth & access control: By default, MCP uses OAuth 2.0; Easy Auth enforces client/identity/tenant restrictions. During setup, App Service authentication must allow unauthenticated requests (the MCP flow still performs OAuth).
Transports: Streamable HTTP works out of the box. SSE additionally requires VNET integration and host.json setting Runtime.Backend.EdgeWorkflowRuntimeTriggerListener.AllowCrossWorkerCommunication=true.
Enablement switch: MCP APIs are enabled by adding extensions.workflow.McpServerEndpoints.enable=true in host.json.
API Center path: preview limitations that matter
When creating MCP servers via API Center backed by Logic Apps, the current preview imposes the following limits:
Fresh Start RequiredYou can't just tack this onto your existing Logic Apps - you'll need to spin up a brand new Standard Logic App resource for each MCP server. No shortcuts here!
Stick to One Connector at a TimeEach MCP server can only work with a single connector type. Need both Outlook and SQL Server? That's two separate servers you'll need to set up.
Play by Microsoft's Rules OnlyThis feature only works with Microsoft's managed connectors - sorry, but your custom connectors and built-in service providers won't make the cut for this party.
Keep It Simple - One Tool, One JobEvery tool you create can handle just one specific action. No multi-purpose tools allowed - think laser-focused rather than Swiss Army knife.
The Reality CheckFor bigger organizations, these limitations can get pretty messy pretty fast. You could end up juggling tons of different MCP servers just to cover your basic workflow needs, which means a lot more planning and management overhead than you might expect.
Why Standard (single-tenant) is the target?
Why Standard Logic Apps Make Sense Here
Standard Gets Its Own Dedicated SpaceThink of Standard as having your own private office building (single-tenant on Azure Functions) where you can set up multiple workflows under one roof. It's like having a whole suite of connected offices rather than renting individual cubicles.
Built for Security-First ScenariosHere's where Standard really shines - it plugs directly into your virtual networks and private endpoints. This is huge when you're letting AI agents poke around your internal systems. You get that extra security layer without jumping through hoops.
Predictable PerformanceWith Standard, you know exactly what kind of performance and response times you're getting. No surprises, no "depends on who else is using the system" moments - just consistent, reliable execution.
The Consumption AlternativeConsumption Logic Apps are more like shared co-working spaces - you're sharing resources with everyone else (multitenant), you get one workflow per app max, and you pay only when things actually run. Great for simple, occasional tasks, but not ideal when you need AI agents to reliably access your private systems with consistent performance.
For MCP servers that need to safely expose internal workflows to AI agents, Standard's architecture just makes way more sense.
Making Your Workflows AI-Friendly
Want AI agents to actually use your workflows correctly? Give them good instructions! Microsoft suggests adding clear descriptions to your triggers, detailed parameter explanations, and marking which fields are absolutely required. Think of it like leaving really good notes for a new team member - the better your documentation, the smarter the AI acts when using your tools.
Your Enterprise Toolkit Just Got Superpowers
Here's the exciting part: you can now expose your entire Logic Apps arsenal - both cloud and on-premises workflows - as tools that AI agents can actually call. Microsoft's throwing around the number "over 1,400 connectors" for a reason. That's your CRM, your databases, your file systems, your communication tools - basically your whole digital ecosystem becomes available to AI assistants.
Keeping Everything Under Control
Don't worry about flying blind once this goes live. You'll have full run history and can plug into Application Insights and Log Analytics for detailed monitoring and audit trails. Testing is pretty smooth too - VS Code lets you quickly validate everything through the "MCP: Add Server" feature, complete with OAuth testing and a list of all available tools. If you're using API Center, you get the added bonus of proper discovery and governance across your entire organization.
Real-World Gotchas (It's Still Preview, After All)
Network Setup HeadachesWant to use Server-Sent Events (SSE)? You'll need both VNET integration and the cross-worker setting enabled, or you're stuck with regular HTTP streaming. Not the end of the world, but something to plan for.
Authentication Can Be TrickyEasy Auth configuration needs to be just right - including that sneaky "allow unauthenticated" toggle. Get it wrong and your OAuth flows will mysteriously fail, leaving everyone scratching their heads.
You're Still the Boss of PerformanceMicrosoft isn't handling throttling, duplicate request handling, or version management for you. These are still your responsibility, but now they matter even more since AI agents will be hitting your endpoints.
Summary
This preview basically turns your Logic Apps into a smart toolkit for AI agents. Your HTTP workflows become OAuth-secured tools, you can organize them properly in API Center, and you can safely connect to private systems through dedicated networking.
If your team is already deep into Logic Apps, this is probably the easiest path to get enterprise-grade AI tooling up and running. Just keep an eye on those API Center limitations, make sure you nail the SSE setup requirements, and don't underestimate the Easy Auth configuration - get those right and you're golden.
Subscribe for updates
Stay updated with the latest news, articles and update directly into your box
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.