Microsoft Onenote

What the Microsoft Onenote MCP server does

The Microsoft Onenote MCP server exposes read access to data in a connected Microsoft 365 / OneNote account. Its tools follow OneNote's hierarchy: notebooks contain sections, and sections contain pages. An agent can start by finding available notebooks, narrow the search to sections or pages, and then retrieve the HTML body of a selected page.

The package does not list tools for creating, editing, moving, or deleting OneNote content. Its documented capabilities are for discovery and reading existing notes.

How it works

The Microsoft Onenote MCP server can be reached in two ways. The hosted option uses the Pipeworx gateway at https://gateway.pipeworx.io/microsoft-onenote/mcp. A client connects to that endpoint using its MCP configuration, while the gateway handles access to the connected Microsoft account.

A local option runs @pipeworx/mcp-microsoft-onenote through npx and communicates over standard input and output. This mode does not require a Pipeworx account or a gateway round trip. The local process exposes only the OneNote pack's tools, while the hosted endpoint also includes shared Pipeworx meta-tools such as ask_pipeworx, discover_tools, search_within, and memory-related tools.

For direct HTTP calls to the hosted gateway, users sign in at Pipeworx, connect Microsoft, and send a Pipeworx key as a bearer token. The gateway also provides tool inspection and tool search endpoints.

Tools and capabilities

• list_notebooks returns notebook IDs, names, creation and modification timestamps, and the default-notebook flag.
• list_sections accepts a notebook ID to inspect one notebook, or no ID to return sections across notebooks. Results include section IDs, names, and timestamps.
• list_pages accepts a section ID for focused browsing, or no section ID to list recent pages across notebooks. Results are newest first and can be filtered with free-text search. Each result includes the page ID, title, and timestamps.
• get_page_content retrieves one page by ID and returns its complete body as HTML.

A typical workflow is to call list_notebooks, choose a notebook, call list_sections, locate pages with list_pages, and then pass a page ID to get_page_content. If the target page is already known, the intermediate discovery steps may not be necessary.

Setup and configuration

For a hosted MCP connection, add the Pipeworx endpoint to an MCP client configuration:

{
  "mcpServers": {
    "microsoft-onenote": {
      "url": "https://gateway.pipeworx.io/microsoft-onenote/mcp"
    }
  }
}

For local stdio mode, use:

{
  "mcpServers": {
    "microsoft-onenote": {
      "command": "npx",
      "args": ["-y", "@pipeworx/mcp-microsoft-onenote"]
    }
  }
}

The README names Claude Desktop, Cursor, and Windsurf as compatible MCP clients. Hosted use requires a connected Microsoft account and a Pipeworx key; local use does not require a gateway account.

Limitations and notes

The Microsoft Onenote MCP server returns page content as HTML rather than a separately described plain-text or structured format. The available pack tools are read-only. Hosted pack endpoints include shared Pipeworx tools in addition to the OneNote tools, so the tool list may be larger than the four pack-specific operations. The local package exposes only the pack-specific operations and does not provide gateway-wide routing through ask_pipeworx.