Azure Wiki Search MCP Server for AI Agents

Overview

This MCP (Model Context Protocol) server exposes Azure wiki content as a searchable knowledge source for AI agents. Implementing the MCP specification, the server provides a small set of tools that let agents discover related wiki pages and retrieve the raw contents of a specific wiki path. Using this server improves agent responses by enabling retrieval of authoritative documentation, troubleshooting notes, and internal guides from the configured Azure wiki project.

The server is designed to be lightweight and easy to run locally or in a developer environment. It integrates with common developer workflows (VS Code, Python virtual environments) and is intended for use as a fast, reliable knowledge retrieval backend for agent frameworks that support MCP.

Features

• Implements the MCP specification so AI agents can call search and retrieval tools.
• Search Edge Wiki for relevant content by query.
• Retrieve full wiki content by providing a wiki path.
• Simple local deployment with uv (ultra-virtual) runner and Python 3.10+.
• Optional VS Code configuration to register the server as an MCP endpoint for agents in-editor.

Installation / Configuration

Prerequisites:

• Python 3.10+
• Git
• VS Code (optional, for integrated development)
• uv runner (used to create and run the virtual environment and launch the server)

Install uv:

• Windows (PowerShell)

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

• macOS / Linux (bash)

curl -LsSf https://astral.sh/uv/install.sh | sh

Clone and set up the project:

git clone https://github.com/coder-linping/azure-wiki-search-server.git
cd azure-wiki-search-server

Create and activate the virtual environment:

• Windows

uv venv
.venv/Scripts/activate

• macOS / Linux

uv venv
source .venv/bin/activate

VS Code integration (optional) Add the following JSON to your User Settings (JSON) or to .vscode/mcp.json in the workspace. Replace and optionally set ORG and PROJECT env vars.

{
  "mcp": {
    "servers": {
      "edge_wiki": {
        "command": "uv",
        "args": [
          "--directory",
          "<absolute path to your cloned folder>",
          "run",
          "src/edge_wiki.py"
        ],
        "env": {
          "ORG": "Your organization, default is microsoft",
          "PROJECT": "Your project, default is Edge"
        }
      }
    }
  }
}

Running the server

• From the activated virtual environment:

uv run src/edge_wiki.py

The server will start and expose MCP-compatible tool endpoints for agent consumption.

Available Tools

The server exposes two primary MCP tools. Below is a quick reference.

Tool name	Purpose	Inputs	Outputs
search_wiki	Find wiki pages related to a free-text query	query (string)	List of search results (title, path, snippet, score)
get_wiki_by_path	Retrieve full wiki content by path	path (string)	Page content (markdown/plain text), metadata

Example tool call payloads (MCP-style JSON examples):

• search_wiki

{
  "tool": "search_wiki",
  "input": {
    "query": "how to configure Edge updates"
  }
}

• get_wiki_by_path

{
  "tool": "get_wiki_by_path",
  "input": {
    "path": "Guides/Edge/Configure-Updates"
  }
}

Responses return structured JSON with the relevant content and metadata suitable for agent consumption.

Use Cases

• Contextual agent responses: An assistant can call search_wiki when a user asks a question about Edge configuration. The agent uses results to ground its answer in the official wiki content and include references or links.
• Document retrieval for multi-step tasks: Agents that need to perform configuration steps can fetch full documentation pages via get_wiki_by_path to present step-by-step instructions or extract commands.
• Debugging and triage: When a user reports an issue, the agent can search the wiki for related incidents, troubleshooting guides, or known issues to reduce diagnostic time.
• Knowledge augmentation: Combine search results with model reasoning to generate explanations, summarize long wiki articles, or produce checklists for engineers.

Concrete example workflow:

1. User: “How do I enable single sign-on for Edge?”
2. Agent: calls search_wiki with query “enable single sign-on Edge”
3. Agent: picks the top result path from search_wiki and calls get_wiki_by_path to fetch the page content.
4. Agent: synthesizes a concise answer with steps and a link to the wiki page.

Notes and Tips

• Configure ORG and PROJECT environment variables to point to the correct Azure wiki repository if not using defaults.
• Use the VS Code MCP configuration to make the server available to in-editor agent plugins like Copilot Chat.
• The server is intended as a retrieval layer; treat results as source material to be cited or summarized rather than authoritative model output.