Microsoft OneDrive MCP Tools for AI Agents
Let an agent search, read, move, share, and delete files and folders in a user's OneDrive.
Available Tools
17 tools available in this integration
MicrosoftOnedrive.CopyItem Copy a file or folder and wait for the copy to finish, returning the new item.
MicrosoftOnedrive.CreateFolder Create a new folder in OneDrive.
MicrosoftOnedrive.CreateShareLink Create a share link for a OneDrive item.
MicrosoftOnedrive.DeleteItem Delete one or more files or folders from OneDrive in a single call.
MicrosoftOnedrive.DownloadFile Download a OneDrive file's content as base64, paging large files by byte offset.
MicrosoftOnedrive.FindDuplicateFiles Find files that are byte-identical copies of each other across folders in one call.
MicrosoftOnedrive.GetCopyStatus Check status of an async copy operation using the token an in-progress copy returned.
MicrosoftOnedrive.GetItem Resolve a single OneDrive file or folder's metadata directly from its id.
MicrosoftOnedrive.GetMyDrive Get metadata about the user's OneDrive (id, name, quota, owner).
MicrosoftOnedrive.GetSharedWithMe List files shared with the current user.
MicrosoftOnedrive.ListFolderItems List files and folders in a OneDrive folder.
MicrosoftOnedrive.ListItemPermissions List who can currently access a OneDrive item: every share link and direct grant on it.
MicrosoftOnedrive.MoveItem Move and/or rename one or more files or folders in OneDrive in a single call.
MicrosoftOnedrive.ReadFile Read the text of a PDF or plain-text file in the user's own OneDrive.
MicrosoftOnedrive.RevokeItemPermission Revoke one or all anonymous sharing grants on a OneDrive item.
MicrosoftOnedrive.SearchItems Search for files and folders in the user's OneDrive.
MicrosoftOnedrive.WhoAmI Identify the current user and confirm OneDrive access for orientation.
Why agent-optimized tools matter
Built for intent, not API schemas
Ask the agent to "find duplicate files in my OneDrive, show me the copies, and leave the originals alone." Arcade picks the right OneDrive actions without guessing item IDs or folder IDs.
Up to 100x fewer tokens
OneDrive searches and folder listings return lightweight file and folder metadata first, while full text or file content is fetched only when asked, so the agent can scan a drive without loading every document into the model.
Complete the workflow end to end
Search a file, inspect its metadata, read or download the content, move or copy it, create a share link, and request approval through one toolkit designed around the complete OneDrive 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 OneDrive
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 PoliciesFor Microsoft OneDrive, you can implement some of these policies:
- Hide MicrosoftOnedrive.DeleteItem from agents unless the roles your IDP marks as OneDrive administrators are active, so an agent cannot remove files or folders outside its assigned scope.
- Restrict MicrosoftOnedrive.RevokeItemPermission to agents operating under your data-owner role, so automation cannot cut off a colleague or external partner's access.
- Deny MicrosoftOnedrive.CreateShareLink to read-only agents, so they cannot publish private files through new anonymous links.
- Require human approval before MicrosoftOnedrive.MoveItem changes the location or name of files in your protected folders, so an agent cannot disrupt an approved folder structure.
- Block MicrosoftOnedrive.CreateShareLink when the target item is marked confidential or the requested scope is anonymous, so private documents cannot be shared without review.
- Require human approval before MicrosoftOnedrive.DeleteItem runs on folders or files outside the agent's temporary workspace, so a mistaken cleanup cannot remove team content.
- Deny MicrosoftOnedrive.DownloadFile outright for files marked confidential, under legal hold, or matching your secrets and credentials patterns, so their contents never reach the model.
- Redact email addresses, sharing-link URLs, and direct-grant identifiers from MicrosoftOnedrive.ListItemPermissions results before the model sees who can access an item.
- Strip personal contact details and external account identifiers from MicrosoftOnedrive.GetSharedWithMe results, so the model can use shared-file metadata without exposing collaborator information.
How to use Microsoft OneDrive with your agent
Pick your MCP client below and follow the setup to connect OneDrive through Arcade. Each end user authorizes their own Microsoft account.
Your questions, answered
A Microsoft OneDrive MCP server exposes OneDrive 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 search, read, organize, share, and manage permitted files and folders.
Microsoft OneDrive does not provide the first-party MCP server used here. Arcade adds the runtime around OneDrive access, including per-user OAuth, agent-optimized actions, approval workflows, response redaction, and scope controls across supported clients.
Arcade agents can identify the current user and drive, search and list files and folders, read text and PDF files, download files, find duplicate files, inspect sharing permissions, list shared items, create folders and share links, copy or move items, rename them, revoke anonymous sharing, and delete files or folders.
Arcade requests Microsoft Graph delegated permissions for the OneDrive actions your workflow enables. Read-only workflows need file and drive read permissions; creating, moving, copying, sharing, revoking access, or deleting requires the corresponding write or file-management permissions. Grant the narrowest scopes that support the workflow.
Yes. Each user authorizes their own Microsoft account through OAuth. OneDrive actions run under that user's authorization, so the agent can access only the drive and permissions that user has approved.
Yes. Contextual Access policies can require human approval before consequential OneDrive actions such as deleting items, moving or copying content, creating share links, or revoking access. The call proceeds only after the policy's conditions have been satisfied.
Arcade proxies OneDrive API calls and does not persist your file content as part of normal tool execution. OAuth credentials are brokered and stored encrypted at rest, and the model never receives raw credentials.
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, OneDrive actions stop until the user reconnects their account.
Microsoft Graph throttling, OneDrive service limits, file-size limits, and tenant or account quotas still apply through Arcade. Large downloads are returned in byte ranges, and callers should handle throttling and retry responses rather than assuming unlimited request volume.