Sign in Book a demo
Back to Tools
Productivity & Docs OAuth
Build Your Agent View Word Documentation
microsoft-word logo

Microsoft Word MCP Tools for AI Agents

Let an agent create Word documents in OneDrive, read .docx content, and append text to existing files.

Available Tools

4 tools available in this integration

MicrosoftWord.CreateDocument

Create a new Word document in OneDrive (4MB upload limit).

MicrosoftWord.GetDocument

Get a Word document's metadata and content (`.docx` only).

MicrosoftWord.InsertTextAtEnd

Append text to the end of a Word document (supports only `.docx`, 4MB limit).

MicrosoftWord.WhoAmI

Get information about the current user and their Microsoft Word environment.

Why agent-optimized tools matter

Built for intent, not API schemas

Ask the agent to "create a project brief, add the decisions from this meeting, and return the document content." Arcade helps it use the right Microsoft Word actions without guessing document IDs or raw API parameters.

Up to 100x fewer tokens

Microsoft Word reads return a document's metadata and content together when the agent opens a file. It can inspect the document it needs without loading unrelated files or a wider OneDrive collection into the model.

Complete the workflow end to end

Identify the user, create a document, read its content, append text, and verify the result through one toolkit designed around the complete document task.

Make actions predictable and controlled

Arcade tools return structured results and useful errors across users and runs. Teams can apply scope restrictions, redaction, and approval policies so agents follow the same controls every time.

Securing Microsoft Word

The more capabilities agents get, the more useful they are. However, more capabilities bring more risks of undesired effects such as data loss and exfiltration.

This is why Arcade provides you with contextual rules that dictate how and when a tool is allowed to run. Arcade’s contextual access lets you govern tool visibility and behavior at every stage of execution and control who can see a tool, what inputs are allowed, and what comes back.

Learn about Contextual Access Policies

For Microsoft Word, you can implement some of these policies:

Access when tools are listed for a user
  • Hide MicrosoftWord.CreateDocument from agents unless the roles your IDP marks as document owners are active, so unauthorized agents cannot create files in users' OneDrive accounts.
  • Restrict MicrosoftWord.InsertTextAtEnd to agents operating under your document-editor role, so a regular user cannot append unreviewed text to shared documents.
Pre-execution before each tool call
  • Require human approval before MicrosoftWord.CreateDocument runs, so an agent cannot create an unsolicited document in a user's OneDrive.
  • Block MicrosoftWord.InsertTextAtEnd when the target document is marked read-only or legally held, so automation cannot alter records that must remain unchanged.
Post-execution after the tool returns
  • Redact the document body MicrosoftWord.GetDocument returns when it contains credentials, access tokens, or personal identification numbers before the content reaches the model.
  • Strip personal profile fields and tenant identifiers from MicrosoftWord.WhoAmI results, so the model receives only the account context needed to choose the right document.

How to use Microsoft Word with your agent

Connect Microsoft Word to your MCP client and give your agent access to OneDrive documents. Follow the walkthrough from authorization to its first document action.

Your questions, answered

A Microsoft Word MCP server exposes Word document capabilities as structured tools that an AI agent can call through the Model Context Protocol. After a user authorizes their Microsoft account, the agent can create, read, and append text to permitted Word documents in OneDrive.

Microsoft does not provide the first-party MCP server used here. Arcade provides the MCP access layer for these Word actions, including per-user OAuth, approval workflows, redaction, and scope controls across supported clients.

Arcade agents can create new Word documents in OneDrive, read metadata and content from .docx files, append text to the end of .docx documents, and identify the authorized user and their Word environment. They cannot edit arbitrary locations in a document or work with non-.docx files through these actions.

Arcade requests the Microsoft OAuth permissions needed for the document actions you enable. Reading documents needs read access, while creating documents or appending text needs write access to the user's OneDrive documents; grant the narrowest permissions that support your workflow.

Yes. Each user authorizes their own Microsoft account through OAuth. Word actions run under that user's authorization, so the agent can access only the documents and permissions that user has approved.

Yes. Contextual Access policies can require human approval before consequential Microsoft Word actions such as creating a document or appending text. The call proceeds only after the policy's conditions have been satisfied.

Arcade proxies Microsoft Word and OneDrive API calls and does not persist your document content as part of normal tool execution. OAuth credentials are brokered and stored encrypted at rest, and the model never receives them.

Any framework or client that can connect to Arcade's MCP gateway or tool APIs works, including LangChain, OpenAI Agents, CrewAI, Claude Desktop, Claude Code, Cursor, Visual Studio Code, Microsoft Copilot Studio, and GitHub Copilot.

Arcade refreshes access while the user's Microsoft authorization remains valid. If access is revoked, rotated, or can no longer be refreshed, document actions stop until the user reconnects their Microsoft account.

Microsoft Graph and OneDrive rate limits and quotas apply to calls made through Arcade. Creating a document and appending text are each limited to 4MB, and the document actions support .docx files only.

Ready to actually ship your agent?

Connect your agent to Microsoft Word with managed OAuth, agent-optimized tools and approval controls for consequential actions.

Build Your Agent Read the Microsoft Word Quickstart