Hi Abhishek,

Mendix has two different MCP servers, and which one you need depends on what you want to do:

Studio Pro MCP Server

• Purpose: Lets AI coding tools build or change your app model
• Runs where: Locally, inside Studio Pro, at development time
• Typical clients: Claude Code, Copilot, Cursor

MCP Server module (Marketplace)

• Purpose: Exposes your running app’s logic to AI clients
• Runs where: Inside the Mendix runtime, in a deployed or live app
• Typical clients: Claude Desktop, other AI agents, other Mendix apps

For a live project with real-time usage, you most likely need Option B, the MCP Server module.

Option A – Studio Pro MCP Server (development time)

Studio Pro MCP Server enables bring your own agent (BYO Agent) capabilities, allowing you to use Maia’s capabilities directly from external clients, including AI coding assistants, agents, and other MCP-based LLM tools. Any changes made through Maia via the MCP Server are reflected in real time within Studio Pro. mendixmendix

Setup:

1. Go to Preferences > AI > MCP Server and select Enable MCP Server. You can also configure the port in this menu. mendix
2. Connect your client. For example, in Claude Code:

   claude mcp add mendix --transport http http://localhost:<port>/mcp

For other clients, such as VS Code Copilot or Cursor, add an HTTP MCP server pointing to http://localhost:<port>/mcp in that client’s MCP configuration.

3. Ask your agent to make changes, for example “create an entity Customer with Name and Email”. Then review the result in Studio Pro and in the Errors pane.

Good to know:

• The MCP Server runs locally on your workstation, so you do not need to be signed in to Studio Pro to use it. mendix
• External tools consume tokens from your own LLM provider account each time they interact with Maia. This doesn’t use Maia Units. mendix
• Recommended practices: Give the agent context. Agent Instructions (AGENTS.md) and Agent Skills (SKILLS.md) follow industry standards, so they work with Maia and other agentic tools. Keep requests small, review each change, and commit after testing. mendix
• This server is only for building the app. It isn’t part of your deployed application.
  
• 
  

Option B – MCP Server module (live / deployed app)

The MCP Server module provides easy low-code capability to set up an MCP (Model Context Protocol) server within a Mendix app. With it you can expose reusable prompts, including the ability to use prompt parameters and list and execute microflow implemented in the application as tools. MCP Server +2

Step 1 – Install the module

Download MCP Server from the Marketplace. It’s already included if you started from the Blank GenAI App template. This page documents MCP Server module versions compatible with Agents Kit 2, so make sure to use an app running on Studio Pro 11.12 or above, which matches your version. mendix

Step 2 – Create the server, for example in an After Startup microflow

Use the Create MCP Server action and set:

• Path, for example mcp-tickets. If Path is set to mcp-ticketsystem, the MCP Client will reach it via the URL https://mymendixapp.com/mcp-ticketsystem/mcp. The Path can’t be mcp, can’t end with /mcp, and can’t contain /. mendix
• ProtocolVersion: The latest version of the MCP Server module only supports v2025-03-26 and the Streamable HTTP transport. mendix
• Authentication microflow, which is strongly recommended. The Input type should be MCPServer and/or System.HttpRequest, to extract required values, such as HttpHeaders, from the request. The return value needs to be a System.User object which represents the user who sent the request. That user is then applied to all tool calls, so your normal entity access rules apply. mendix
  
• 
  

Step 3 – Expose microflows as tools

After creating the server, use Add Tool for each microflow you want to expose, for example Tool_GetOpenTickets. Rules:

• Input: primitive parameters and/or a MCPServer.Tool object. If you don’t pass a schema, one is generated from the microflow parameters.
• Return value: The return value must be either of type String or TextContent. Return TextContent if you want to set IsError = true for handled errors. mendix
• Give each tool a clear name and description, because the LLM uses these to decide when to call it.
  
• 
  

Step 4 – Optionally add prompts

Use Add Prompt, with Populate Prompt Argument List for prompt parameters, to publish reusable prompt templates.

Step 5 – Deploy and connect a client

Deploy the app, then point your MCP client to https://<your-app-url>/<Path>/mcp. For example, in Claude Code:

claude mcp add my-mendix-app --transport http https://myapp.mendixcloud.com/mcp-tickets/mcp --header "Authorization: Basic <base64-credentials>"

Step 6 – Use the examples

• The module has an Example Implementations folder with a server, an authentication microflow, a tool, and a prompt.
• The GenAI Showcase App on the Marketplace has a full working example. See the microflow ACT_MCPServerConfiguration_InitializeMCPServer.
• Agent Editor offers a simpler, configuration-based way to publish an MCP service.
  
• 
  

Production tips

• Don’t run it without authentication on a public URL. If no authentication is enabled for the MCP Server, it can be accessed by any service without being authorized specifically. mendix
• For tools that change data (send emails, create orders, delete records), keep a human in the loop with confirmation steps.
• Expose only small, purpose-built microflows, not general-purpose CRUD.
• If you get a Java library conflict after importing the module, run App > Synchronize dependencies and restart.
• The module is still evolving, so check the release notes before upgrading.
  
• 
  

In short: use Option A if you want an AI coding assistant to help you build the app. Use Option B if you want AI agents to call your live app’s business logic.

Hope this helps! If this solution works for you, kindly mark it as the accepted answer so it can help others facing the same issue.

Sources:

• Studio Pro MCP Server – Mendix Docs
• MCP Server module (Agents Kit 2) – Mendix Docs
• Maia MCP Client – Mendix Docs

Hi Abhishek,

The implementation depends on whether you want an AI assistant to help develop your Mendix application or interact with a running application.

1. For AI-assisted development in Studio Pro

If you are using Studio Pro 11.10 or later:

1. Sign in to Studio Pro and open your application.
2. Go to Preferences → AI → MCP Server.
3. Enable MCP Server and note the configured port (default: 7782).
4. Configure your MCP-compatible AI client to connect to:
5. http://localhost:7782/mcp
6. For example, with Claude Code, run:
7. claude mcp add mendix --transport http http://localhost:7782/mcp
8. Test with a read-only request first, then review any model changes made by the AI in Studio Pro before committing them.

This server runs locally while Studio Pro is open; it is not a runtime endpoint for your deployed application.

2. For AI access to a running Mendix application

If your goal is to let an external AI assistant perform business operations, use the Mendix MCP Server module:

1. Install the compatible MCP Server module in your Mendix application.
2. Create an MCP server and configure its path and protocol version.
3. Select specific microflows to expose as MCP tools, with clearly defined input schemas.
4. Apply authentication and authorization, and ensure each exposed microflow validates its inputs and the user's permissions.
5. Deploy the application and configure your MCP client to connect to the published endpoint. Test the connection and each tool in a non-production environment before enabling it in production.

For example, you could expose a microflow that retrieves a support ticket by ID. Avoid exposing unrestricted create, update, or delete operations without appropriate authorization and validation.

My recommendation: For a real-time AI assistant that needs to access business data or trigger operations in your live application, start with the Mendix MCP Server module. Expose only the business operations the assistant actually needs.

Official documentation:

• Studio Pro MCP Server
• Mendix MCP Server Module

Hope this helps!

Hi Subhash, Guatham

Thanks for your informative answers, I was wondering if there is any way by which instead of tools MCP/Agent reads whole of our data and then answers questions based on our input.

If theres any sort of way with which it can be achieved please do let me know.