Onedrive

What the Onedrive MCP server does

The Onedrive MCP server gives an MCP-compatible AI client access to selected OneDrive and Microsoft 365 file operations. It can browse folders, search across file names and content, inspect individual items, retrieve text from files, find items shared with the connected user, and report drive profile information.

The server works against the connected Microsoft account. Results include identifiers and practical file details such as names, sizes, folder status, modification times, web URLs, and parent references. This makes it suitable for agents that need to locate documents, inspect their metadata, or read supported text files before answering a question.

How it works

Two connection modes are documented. The hosted option uses a Pipeworx MCP endpoint at https://gateway.pipeworx.io/onedrive/mcp. Before using that endpoint, sign in to Pipeworx and connect a Microsoft account. The same gateway can also be accessed through its full endpoint, although the full gateway lists tools from all available packs directly.

A local option runs @pipeworx/mcp-onedrive with npx and communicates through standard input and output. The local process exposes only the OneDrive pack's tools, while the hosted pack endpoint also includes Pipeworx-wide meta-tools such as routing and discovery tools. The hosted connection can therefore route questions beyond this pack through ask_pipeworx.

Setup and configuration

For a hosted MCP client, add the pack URL as an MCP server. A local setup uses the following command and does not require a Pipeworx account:

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

The package can also be started directly with npx -y @pipeworx/mcp-onedrive to confirm that it launches. For users without an MCP client, Pipeworx documents HTTP access to individual tools using a bearer key. The key is obtained from a Pipeworx account after connecting Microsoft.

Tools and capabilities

The Onedrive MCP server provides these documented tools:

• onedrive_list_files lists items in the root or a folder path relative to the drive root.
• onedrive_search_files searches file names and content using a query string.
• onedrive_get_file returns metadata for a file or folder identified by item ID.
• onedrive_get_file_content returns text content for a file by item ID.
• onedrive_list_shared lists items in the user's “Shared with me” collection and identifies the sharer.
• onedrive_get_profile reports drive type, quota usage, total quota, and owner display name.

Content retrieval is capped at approximately 100,000 characters and indicates when output is truncated. The tool is intended for plain-text, Markdown, and CSV files. Office documents, PDFs, images, and other binary formats may return unreadable bytes rather than parsed document content.

Limitations and notes

Hosted pack connections include shared Pipeworx meta-tools in addition to the six OneDrive tools. The exact hosted tool scope can change; the README recommends treating the connection's initialize response as authoritative. The local stdio process does not include those shared tools or cross-pack routing.

The documented operations focus on listing, searching, reading, metadata, sharing, and profile information. Use the item ID returned by listing or searching when inspecting an individual file, and expect text output limits when reading content.