Your AI coding agent, fluent in Mendix, right inside Studio Pro

If you have used Claude Code, Codex or GitHub Copilot CLI on a regular codebase, you know the feeling. You describe what you want, the agent reads the project, makes the change, runs the tests, and comes back with a result. Mendix deserves that same loop, but a Mendix app is not a folder of source files. The model lives in Studio Pro, and so the agent has to work where the model lives: through Studio Pro’s own tools, with Studio Pro as the source of truth.

C'est ce que Concord2 Terminal does. It runs a full CLI coding agent inside Mendix Studio Pro, grounded in the project you already have open. Pick the agent you already use — Claude Code, Codex or GitHub Copilot CLI, with Grok Build as an experimental option — and Concord gives it a dockable terminal next to your model, connects it to Mendix-aware context and Studio Pro’s own actions, and keeps you in control: you review what it changes, and you decide how much it may do without asking.

It is also a preview of something bigger. Concord is the preview of Mendix’s upcoming Bring Your Own Agent (BYOA) capability, now in development at Mendix R&D. Published on the Marketplace by the Mendix CoE Team, it is how you can work this way today, with more than 3,700 downloads so far.

Avant de commencer

Six things worth knowing before you dive in.

1. Versions

Concord 7.5.2 supports Studio Pro 11.10 through 11.16, on Windows x64 and macOS. Studio Pro 10.x is not supported. On macOS, Studio Pro 11.13 and later need Apple silicon; Intel Macs are supported through 11.12.

2. Your agent, your subscription

You bring a signed-in CLI. Claude Code is the fully verified reference CLI; Codex and GitHub Copilot CLI have verified connections; Grok Build is project-wired, with deeper workflow coverage still maturing. Any other terminal-based agent will run in a Concord tab, but only those four are configured for you. Concord adds no subscription and no API key of its own.

3. Studio Pro’s MCP server is the key setting

In Modifier > Préférences, activez le Serveur MCP (you’ll find it in the Maia or AI section, depending on your version), choose a free port (give each Studio Pro you keep open its own) and save. Concord reads that port from Studio Pro’s own settings; it never guesses one.

4. Maia est facultatif

Concord works alongside Maia, but it does not need Maia running. For the rare step that neither Studio Pro’s MCP server nor Concord can do directly, your agent may hand the step to Maia (usually by giving you a prompt to paste into Maia’s chat) and you will see it when it does.

5. Restart fully

After installing or upgrading Concord, enable Développement d'extensions. Ouvert Extensions > Concord > Open Concord2 Terminal once, then close Studio Pro completely and reopen it before you start an agent. After an upgrade, the Concord menu and pane will not come back until you do.

6. Make friends with Setup Status

In the Concord pane, click the Paramètres gear, choose View setup status, puis Run status check. It checks the connection to Studio Pro’s MCP server, your CLI configuration, and a few optional extras such as Playwright, and offers a Fixer button where it can. Extension Development and the MCP server are Studio Pro preferences you set once per installation; the module is added to each app, and opening the pane once prepares that app.

What you can ask it to do

Read your app

Ask it to explain a microflow, a page or your whole customer-onboarding flow; find where an entity is used; read the runtime log after a failed run. It works from the project you have open in Studio Pro, not from an export — just save your work first, so unsaved edits are included.

Demander Mendix le Career Centre

“How do I add a scheduled event?”

“What does this validation rule do?”

It looks the answer up in Mendix’s own documentation through Studio Pro, cites where it came from, and tells you when it couldn’t verify an answer.

Change your model

Ask for a new entity with its attributes, an extra attribute on an existing one, a microflow call, a constant — or bigger pieces such as new microflows and pages. Every change is made through Studio Pro itself, so what you see in Studio Pro is what actually changed. Newer Studio Pro versions let the agent build more.

Bring a prototype to life

Give it a prototype — a screenshot, a mock-up, an existing HTML page — and ask for the Mendix pages. It is remarkably good at translating a design into pages, layouts and widgets that follow your app’s conventions, and with Playwright set up it checks its own result in the browser.

Run and fix

Ask it to run the app and fix what breaks. It starts or restarts the app, waits until it is actually up, reads the console and build errors, and keeps going. With Playwright set up, it can open the running app in a browser and check the pages it just built. (Automatic run is Windows-only today; on macOS it asks you to press Run.)

Security, too

Ask it to set up security: the security level, user and module roles, page and microflow access, entity access rules. It does this through Studio Pro’s own security screens, so you can watch it happen. Review the result line by line before you deploy. (On macOS, Setup Status will ask for the Accessibilité permission first.)

Upgrade modules and whole apps

Ask it to update a Marketplace module and clean up what breaks, or to carry an app from an older Mendix version to the current one. On a big jump — say, Mendix Entre 9 et Mendix 11 — it does much of the heavy lifting: reading the errors, updating widgets and modules, fixing what it can, and handing you a clear list of what still needs a human. Some steps stay manual, but the upgrade goes a great deal faster.

Partir de zéro

Blank app? Quick Start lets you pick a look and an app idea and hands you a single prompt for Codex or Claude Code. Press Entrer and watch the build happen in the terminal, stepping in whenever you want. Sign-in and user roles are included by default.

What changed with Studio Pro 11.15 and 11.16

Studio Pro has had its own MCP server since 11.10, and 11.16 made it broader. Alongside model editing, knowledge-base search and Marketplace installs, an outside agent can now also start and stop the app, read the runtime log, translate texts, and commit or push.

Concord does not compete with that. It sends your agent to Studio Pro’s own MCP server first wherever it covers the job — and on 11.16 that is most of the job — then fills in the rest: a terminal inside Studio Pro, your CLI wired up to the project, grounded Mendix answers, run-and-fix, project memory, and safety rules. For the few things neither can do yet, such as nanoflows and rules, the agent tells you so and points you to the route that works.

One thing Concord keeps in your hands: since 11.15, Studio Pro lets outside agents commit and push, and Concord tells every agent to leave that to you.

If you are already using Concord, update to 7.5.2 whichever Studio Pro you are on: update the module in place from the Marketplace (don’t delete it first), restart Studio Pro fully, and start a fresh agent terminal.

Working safely with an AI agent

Commit or back up first

Before you ask any agent to change your app, commit it or take a backup. Undo helps, but it does not cover everything an agent does — some edits are saved straight to disk, and some fixes clear the undo history. The commit is your real safety net.

Review everything; keep only what you understand

Treat the agent’s work like a pull request from a fast, confident colleague.

Commit and push commands stay yours

Concord tells every agent never to commit or push, and never to delete or rewrite existing parts of your model without your explicit go-ahead; Concord’s own tools refuse destructive changes outright. These are instructions the agent follows — the approval settings in your own CLI are what enforce them, so keep those where you want them.

Some automation is Windows-only today

Reading Studio Pro’s Errors pane, answering a Studio Pro dialog, fixing outdated widgets and restarting a stuck agent session work on Windows; on macOS the agent tells you what to do instead.

It runs on your machine

Concord has no server of its own and sends no telemetry. Your prompts, and whatever the agent reads from your model, go to your CLI’s provider under your own subscription — the same as using that CLI anywhere else. The one optional credential is a Mendix personal access token for Marketplace module updates, stored in Windows Credential Manager or the macOS Keychain. Concord adds a few configuration files to your project; one of them holds a project-specific token, so keep it within your team.

Questions, feedback and help

The best place for questions and feedback is the #concord channel in the Mendix Community Slack — that is where we hang out, and what you report there is exactly what shapes the next release.

For setup problems, run État de la configuration d'abord (Paramètres > View setup status > Run status check); la Documentation onglet sur le Marketplace listing has the details. Concord is self-supported by the Mendix CoE Team and is not covered by Mendix Support produit.

• Construit par le Mendix CoE Team — Ricardo Perdigao · Kelly Seale.
• Merci à Mendix R&D for the platform foundations — Studio Pro, Studio Pro MCP and Maia — that Concord works alongside.

Lire la suite

1. Concord2 Terminal on the Mendix Marketplace
2. Studio Pro MCP server documentation
3. Mendix Community Slack, #concord