Array
(
)
Contact sales

Connecting MTA to your AI assistant with the Model Context Protocol (MCP)

mcp-protocol.

Why we chose an open standard instead of building a chatbot

Giving your existing AI tools access to your test environment

If you’ve used Menditect Test Automation (MTA) before, you know what the core engine does well: it runs microflows, handles assertions, and rolls back database changes so you don’t end up with hundreds of fake records in your development database.

Up until MTA 3.1, though, setting up those tests was completely manual. You had to open the web interface, create a suite, add steps, select entity types, and type in parameter bindings and expected values by hand. It worked, but when you had 30 or 40 microflows to cover, it took a lot of tedious clicking.

With the upcoming MTA 3.2 release, we wanted to make that process much faster. But instead of building another standalone AI chatbot with its own subscription, we decided to support the Model Context Protocol (MCP).

MCP is an open standard (started by Anthropic and adopted across tools like Cursor, Claude Code, Gemini Antigravity, and GitHub Copilot) that lets AI assistants talk directly to external developer tools. By exposing MTA’s capabilities as an MCP server, your AI assistant can now create and configure tests inside your existing MTA instance.

The real benefit: less clicking, better test coverage

Helping teams catch up on test debt in Mendix 9 and 10 apps

Let’s be honest: writing test cases by hand in any tool is rarely anyone’s favorite part of a sprint. When project deadlines get tight, automated testing is often the first thing pushed to the backlog.

Over time, especially in mature applications running on Mendix 9 or 10, that creates a lot of test debt. Teams end up with large, business-critical apps that have almost no automated tests. That makes refactoring scary, and it makes planning an upgrade to Mendix 10 or 11 feel like walking through a minefield.

With MTA 3.2, you can use your AI assistant to do the heavy lifting:

  • The AI handles the setup: You prompt your assistant to look at a microflow. It inspects the logic, plans the test scenarios, and calls MTA’s MCP tools to set up the test cases, parameter bindings, and assertions.
  • Covers all test levels: You can set up unit tests for isolated calculation microflows, component tests for domain rules, end-to-end backend process tests across multiple microflows, and UI tests for web forms.
  • Automated data seeding and cleanup: The assistant can configure test steps to create the necessary test objects beforehand. Because MTA rolls back database transactions after each test, your environments stay clean.
  • Works when documentation is missing: If you have user stories or Jira specs, you can include them in your prompt. But if those are outdated or missing, the assistant can inspect the Mendix model directly to see the real logic and decision branches.
See MTA MCP in action!

How it works behind the scenes

Connecting the Mendix model, your AI, and MTA

The workflow is straightforward and keeps you in control:

  1. Model inspection (mxcli or Studio Pro MCP): The AI reads your application structure in a safe, read-only manner using the Mendix CLI (mxcli), or via the native Studio Pro MCP server available in Mendix 11.12 and higher. It sees microflow activities, parameters, and domain entities.
  2. Menditect agentic skills: Menditect does not only deliver MCP tools, but also ready-to-use agentic test skills via our public GitHub repository (Menditect/agentic-test-skills) and the Mendix Marketplace. These skills guide your AI assistant with testing best practices, naming standards, and workflow rules.
  3. Review the plan: The AI proposes a test plan showing the scenarios it found (happy paths, boundary values, error branches). You review it and make tweaks if needed.
  4. MTA MCP server: Once you confirm, the AI calls MTA’s MCP tools to create the test suite, test cases, and variations on your MTA server.

To help you get started quickly, Menditect also publishes an open source template for workspace configuration with MTA MCP, MTA plugin MCP, and mxcli or Studio Pro MCP in the repository Menditect/agentic-test-tools.

architecture-schema

Practical cost management: generate once, run forever

Avoiding unnecessary token costs in your test pipelines

A common concern with AI in software development is unpredictable API costs.

With MTA 3.2, you only use AI “once”, during development, to analyze the microflow and generate the test steps. Once those tests are saved in MTA, they run natively in your daily CI/CD pipeline or before releases at compiled speed, with “zero AI token costs.”

You can also keep generation costs low by using a stronger reasoning model (like Claude Sonnet or Gemini Pro) to design the test plan, and then letting a faster, cheaper model (like Gemini Flash or Claude Haiku) execute the tool calls to configure the steps in MTA.

What this means for your team

  • For developers: Less time spent clicking through forms to set up tests, and more confidence that your microflows actually handle edge cases.
  • For testers: Easy to build multi-variation test matrices across unit, process, and UI levels without getting bogged down in boilerplate setup.
  • For team leads and product owners: A realistic way to get good test coverage on legacy Mendix 9 and 10 applications before starting major upgrades.
Schedule free trial intake

Mendix builds the future; Menditect ensures it works.

MTA 3.2 is coming out in a few weeks. If you want to see how it works on real Mendix microflows, check out these sessions:

See MTA in action:

Dick van Gorkum

Get in touch with Menditect

Fill out this form and we answer your question or help you out.

Footer contact ENG

  • Menditect
  • Databankweg 12G
  • 3821 AL Amersfoort
    The Netherlands
info@menditect.com
+3185 - 4010494
  KvK: 82975795
   
Follow us on LinkedIn
Privacy statement End user licence agreement Service Level Agreement
Design & development: TeamCreative

Menu

Plan your meeting with sales directly

Enter your information and choose a day and time when you would like to meet Menditect sales (1,5 hrs).