Add documentation for Maia MCP Client (Studio Pro 11.8+)

content/en/docs/refguide/modeling/mendix-ai-assistance/_index.md

@@ -46,6 +46,10 @@ Guidance:
 * **Maia Learn** – helps you to quickly learn Mendix core concepts and get started with Studio Pro. For more information, see [Maia Learn](/refguide/maia-learn/).
 * **Maia Explain** – helps you easily understand a microflow or a nanoflow. It explains the general purpose of the logic and highlights specific technical details to help you understand the logic further. For more information, see [Maia Explain](/refguide/maia-explain/).
 
+Extensibility:
+
+* **MCP Client** – helps you connect external [MCP](https://modelcontextprotocol.io/introduction) servers to Maia, giving it access to third-party tools during chat. For more information, see [MCP Client](/refguide/maia-mcp/).
+
 Recommenders:
 
 * **Best Practice Recommender** – helps you inspect your app against Mendix development best practice detecting and pinpointing development anti-patterns and, in some cases, automatically fixing them. For more information, see [Best Practice Recommender](/refguide/best-practice-recommender/).
@@ -96,6 +100,7 @@ The table below presents all the third-party services each Maia capability uses
 | Maia for Workflows | [Claude in Amazon Bedrock](https://aws.amazon.com/bedrock/claude/) | Project context to fulfill user prompt request |
 | Translation Generator | [Amazon Translate](https://aws.amazon.com/translate/) | All translatable texts in the application, for example, labels, button names, and menu items |
 | Validation Assist | No third-party services used | NA |
+| MCP Client | User-configured external MCP servers | Data sent depends on the tools provided by the connected server. Maia sends tool inputs as determined by the LLM based on your chat messages. |
 | Maia Rewrite | [Llama 3.1 8B](https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/MODEL_CARD.md) hosted in Mendix AWS environment | The draft question description from users |
 | Maia Summarize | [Llama 3.1 8B](https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/MODEL_CARD.md) hosted in Mendix AWS environment | [Community](https://community.mendix.com/p/community) threads |
 | Maia Create User Story | [Llama 3.1 8B](https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/MODEL_CARD.md) hosted in Mendix AWS environment | User prompts |

content/en/docs/refguide/modeling/mendix-ai-assistance/maia-mcp.md

@@ -0,0 +1,130 @@
+---
+title: "Maia MCP Client"
+linktitle: "MCP Client"
+url: /refguide/maia-mcp/
+weight: 8
+description: "Describes the features in Maia MCP Client."
+#If moving or renaming this doc file, implement a temporary redirect and let the respective team know they should update the URL in the product. See Mapping to Products for more details.
+---
+
+## Introduction
+
+{{% alert color="info" %}}
+To use Maia MCP Client, an internet connection and signing in to Studio Pro are required.
+{{% /alert %}}
+
+Maia MCP Client, available from Studio Pro 11.8, helps you connect Maia to external [Model Context Protocol (MCP)](https://modelcontextprotocol.io/introduction) servers. Once connected, Maia can call tools provided by these servers during chat — for example, querying an external API, looking up design assets, or interacting with third-party services. You can find available MCP servers on the [MCP servers directory](https://modelcontextprotocol.io/examples). 
+
+For information on the current limitations, see the [Limitations](#limitations) section below.
+
+## Using MCP Client {#using-mcp-client}
+
+To open the MCP Client panel, click the **Configure MCP Connections** ({{% icon name="plug" %}}) icon in the **Maia** toolbar.
+
+### Adding an MCP Server {#adding-server}
+
+To add a server, fill in the following fields and click **Add**:
+
+* **Server Name** — A display name. Must be unique among your configured servers.
+* **URL** — The server endpoint. Must use `https://` or `http://`.
+* **Connection type** — **HTTP (Streamable)** for most servers, or **SSE (Legacy)** for older servers that use Server-Sent Events instead of Streamable HTTP.
+* **Authentication** — See the [Authentication](#authentication) section below.
+
+#### Authentication {#authentication}
+
+Choose how Studio Pro authenticates with the MCP server:
+
+* **No Auth** — select this for servers that do not require credentials.
+* **Bearer Token** — paste an API token or personal access token. The token is stored securely and is not exposed after entry.
+* **OAuth 2.1** — a browser window opens so you can authorize Studio Pro with the server's provider. Two optional fields appear:
+    * **Client ID** — required if the server does not support automatic client registration.
+    * **Client Secret** — required by some providers (for example, GitHub).
+
+Most major providers — including GitHub and Figma — do not support automatic client registration. For these, you must first create an OAuth application in the provider's developer portal and then enter the **Client ID** (and **Client Secret** if required).
+
+##### Example: Connecting to GitHub Copilot MCP {#github-example}
+
+1. Go to [GitHub Developer Settings](https://github.com/settings/developers) and click **New OAuth App**.
+2. Fill in the following information:
+    * **Application name** — for example, `Mendix Studio Pro`
+    * **Homepage URL** — for example, `https://mendix.com`
+    * **Authorization callback URL** — `http://localhost:{port}/mcp/oauth/callback` (see the [Callback URL](#callback-url) section below)
+3. Click **Register application**, then copy the **Client ID**.
+4. Click **Generate a new client secret** and copy it — GitHub shows it only once.
+5. In Studio Pro, add a server with the following details:
+    * **URL** — `https://api.githubcopilot.com/mcp/`
+    * **Authentication** — **OAuth 2.1**
+    * **Client ID** and **Client Secret** — from the steps above
+6. Click **Add**, authorize in the browser, and the connection completes.
+
+{{% alert color="info" %}}
+For Figma MCP, the simplest option is to generate a personal access token in Figma and use **Bearer Token** authentication. If you need OAuth, register an app at [Figma Developers](https://www.figma.com/developers), add the callback URL, and request the `mcp:connect` scope.
+{{% /alert %}}
+
+##### Callback URL {#callback-url}
+
+The OAuth callback URL format is `http://localhost:{port}/mcp/oauth/callback`. The port is assigned by Studio Pro's built-in web server. You can find the actual port within the callback URL that Studio Pro generates when you add the server.
+
+{{% alert color="info" %}}
+If your OAuth app requires a fixed callback URL, you can start Studio Pro with the `--web-server-port=45678` command-line flag to use a specific port.
+{{% /alert %}}
+
+### Managing Connections {#managing-connections}
+
+Each server appears as a card in the MCP Client panel:
+
+* **Enable or disable** a server using the checkbox. Disabling a server disconnects it and removes its tools from Maia.
+* **Expand** the card to see its tools.
+* **Remove** a server from the expanded view.
+
+#### Connection Status {#connection-status}
+
+| Color | Meaning |
+| --- | --- |
+| Green | Connected, tools available. |
+| Orange | Connecting, authenticating, or disconnecting. |
+| Blue | Loading tools from the server. |
+| Red | Error. Check the error message for details. |
+| Grey | Disconnected or disabled. |
+
+If a connection drops, Studio Pro reconnects automatically. During reconnection, the status shows the retry progress (for example, **Retrying (2/5)...**). If all retries fail, disable and re-enable the server to try again.
+
+### Managing Tools {#managing-tools}
+
+After connecting, the server's tools appear in the expanded card. Each tool shows a name and description from the server.
+
+Tools from a new server are **disabled by default**. Select a tool's checkbox to enable it. Use the **select all** checkbox to enable or disable all tools at once. Only enabled tools are available to Maia.
+
+### How Maia Uses MCP Tools {#using-tools}
+
+During chat, Maia considers all enabled MCP tools alongside its built-in capabilities. When Maia calls a tool:
+
+1. The chat shows **Executing {tool} from {server}...** with a spinner.
+2. The tool runs on the MCP server and returns a result.
+3. Maia uses the result to formulate its answer.
+
+You see the answer from Maia, not the raw tool output.
+
+## Troubleshooting {#troubleshooting}
+
+| Problem | Solution |
+| --- | --- |
+| "This server requires manual OAuth client registration" | Register an OAuth app with the provider and enter the Client ID. See the [GitHub example](#github-example) given above. |
+| "Token exchange failed" or "incorrect_client_credentials" | Verify your Client ID and Client Secret. Check that the callback URL matches. |
+| Authentication stuck in "Authenticating..." | Check that no firewall blocks localhost. Remove the server and add it again to restart the flow. |
+| Connection fails immediately | Verify the URL. Try switching between **HTTP (Streamable)** and **SSE (Legacy)**. |
+
+## Limitations {#limitations}
+
+Maia MCP Client has the following limitations:
+
+* Only MCP tools are supported. Other MCP capabilities, such as resources and prompts, are not available.
+* Only remote servers (HTTP or SSE) are supported. Local stdio servers are not supported.
+* Tools from new servers are disabled by default.
+* OAuth requires localhost access for the [callback URL](#callback-url).
+* Server configurations are stored per project and are not shared across projects.
+
+## Read More
+
+* [Mendix AI Assistance (Maia)](/refguide/mendix-ai-assistance/)
+* [Maia Chat](/refguide/maia-chat/)