Access the Ubuntu wiki with an MCP server

Testing

This page has been identified as needing testing.

An MCP server is available that can be used to read and query content from the Ubuntu wiki.

In this guide, you will set up the MCP on Ubuntu using the following tools:

• Inference snap: an easy way to install an AI model that is optimized for your machine.
• OpenCode: an open-source, model-agnostic AI coding harness.
• MediaWiki MCP Server: an open-source MCP server that can run locally on your device.

Note
This guide only describes a local, read-only configuration. It does not provide credentials or access to perform edits on the Ubuntu Wiki using the MCP.

Install OpenCode

There are a number of methods for installing OpenCode.

Here, we will assume you are using NPM:

npm i -g opencode-ai

Configure OpenCode

Create or edit the configuration file at ~/.config/opencode/opencode.jsonc, remembering to replace <username>:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "mediawiki-mcp-server": {
      "type": "local",
      "command": ["npx", "-y", "@professional-wiki/mediawiki-mcp-server@latest"],
      "environment": {
        "CONFIG": "/home/<username>/.config/mediawiki-mcp/config.json"
      }
    }
  }
}

This ensures that the MCP server will be installed and available to OpenCode.

Configure the MCP server

Next, configure the MCP server itself:

{
  "defaultWiki": "wiki.staging.ubuntu.com",
  "wikis": {
    "wiki.staging.ubuntu.com": {
      "sitename": "Ubuntu Wiki",
      "server": "https://wiki.staging.ubuntu.com",
      "articlepath": "/$1",
      "scriptpath": "/w",
      "readOnly": true
    }
  }
}

Install the inference snap

For this guide, you need to install a model with tool support.

We will use GLM 4.7 Flash but you can choose any model you prefer:

sudo snap install glm-4-7-flash

Check the output of glm-4-7-flash status:

engine: cpu
services:
    server: active
    server-webui: active
endpoints:
    openai: http://127.0.0.1:8354/v1
    webui: http://127.0.0.1:8355/
model:
    name: glm-4.7-flash

Note the openai endpoint and the model name, then add those details to the MCP configuration:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "mediawiki-mcp-server": {
      "type": "local",
      "command": ["npx", "-y", "@professional-wiki/mediawiki-mcp-server@latest"],
      "environment": {
        "CONFIG": "~/.config/mediawiki-mcp/config.json"
      }
    }
-  }
+  },
+  "provider": {
+    "glm47flash": {
+      "api": "openai",
+      "name": "GLM 4.7 Flash Inference Snap",
+      "options": {
+        "baseURL": "http://127.0.0.1:8354/v1"
+      },
+      "models": {
+        "glm-4.7-flash": {}
+      }
+    }
  }
}

Chat with the wiki

Run opencode.

Type /models in OpenCode and select glm-4.7-flash.

Ask a question about the wiki, such as:

List the last 5 new pages in the wiki.

The last five pages that were created should be returned, which you can check against Special:NewPages. Example output:

List the last 5 new pages in the wiki
+ Thought: 541ms
mediawiki-mcp-server_get-recent-changes [limit=5]
+ Thought: 1.6s
The 5 most recently created pages:
1. User talk:Mrbite — 2026-09-09 (Mrbite)
2. User:Shanecrowley/drafts/AI — 2026-09-08 (Shanecrowley)
3. Install the Arduino IDE — 2026-09-08 (Marek-suchanek)
4. Ubuntu Wiki:Migration/Temporary/DesktopTeam/TestPlans/Localsearch — 2026-09-07 (Aleasto)
5. Configure SSH to use two-factor authentication — 2026-09-07 (Marek-suchanek)

Next steps

Refer to the README in the GitHub repo to learn about what else is possible with the MCP server.