SharePoint (Microsoft 365)

SharePoint

This connector works with SharePoint Online through Microsoft Graph v1.0 (https://graph.microsoft.com/v1.0): find sites, browse and search document libraries, read file details and who has access, read and write list items, create lists, folders and small text files, and read modern pages. It works with Microsoft 365 work or school accounts (personal Microsoft accounts have no SharePoint). Everything runs as the user who authorized the connector: the AI sees and changes exactly what that user can in SharePoint.

Setup (one-time, about 10 minutes)

1. Open https://entra.microsoft.com (or https://portal.azure.com) → Applications → App registrations → New registration.
2. Name it (e.g. "AnythingMCP Microsoft 365") and choose the supported account types:
   • Accounts in this organizational directory only (single tenant): only your organisation's users can sign in. Set MICROSOFT_TENANT_ID to the Directory (tenant) ID shown on the app's Overview page (a GUID), or to your domain such as contoso.onmicrosoft.com.
   • Accounts in any organizational directory (multi-tenant): set MICROSOFT_TENANT_ID to organizations.
   • ... and personal Microsoft accounts: set MICROSOFT_TENANT_ID to common (work, school and personal accounts) or consumers (personal accounts only). The value must match the app's account type, otherwise Microsoft refuses the sign-in (AADSTS50194 or AADSTS50020).
3. Under Redirect URI pick platform Web and enter https://cloud.anythingmcp.com/api/mcp-oauth/callback on AnythingMCP Cloud, or <your AnythingMCP URL>/api/mcp-oauth/callback when you host AnythingMCP yourself. It must match character for character (AADSTS50011 otherwise).
4. API permissions → Add a permission → Microsoft Graph → Delegated permissions: add offline_access, User.Read, Sites.ReadWrite.All, Files.ReadWrite.All.
5. Certificates & secrets → New client secret. Copy the secret Value (not the Secret ID); it is shown once. Note its expiry date: when it expires the connector stops working until you paste a new one.
6. Paste the Application (client) ID into MICROSOFT_CLIENT_ID, the secret value into MICROSOFT_CLIENT_SECRET, the tenant value from step 2 into MICROSOFT_TENANT_ID, and install.
7. On the connector page click Authorize with Provider and sign in with the account whose SharePoint sites the AI should work with. The refresh token is stored encrypted and access tokens (about 1 hour) are renewed automatically.

One Entra app for all Microsoft connectors: the app registration made for the Outlook connector works here as well. Open it, add the delegated permissions from step 4 under API permissions, keep the same redirect URI, and reuse the same MICROSOFT_CLIENT_ID, MICROSOFT_CLIENT_SECRET and MICROSOFT_TENANT_ID. Each connector asks only for its own scopes when you authorize it. The permissions per connector:

• Outlook: Mail.ReadWrite, Mail.Send, Calendars.ReadWrite
• OneDrive: Files.ReadWrite
• Excel: Files.ReadWrite (plus Files.ReadWrite.All for workbooks in SharePoint or shared with the user)
• SharePoint: Sites.ReadWrite.All and Files.ReadWrite.All Every connector also needs offline_access and User.Read.

Admin consent: these delegated permissions normally do not need an administrator, but many organisations turn off user consent. If the sign-in shows "Need admin approval" (AADSTS65001 or AADSTS90094), a tenant administrator opens the app registration → API permissions → Grant admin consent for <organisation> once; afterwards every user can authorize. Each connector still works only with what the account that signed in may see.

Scopes: offline_access User.Read Sites.ReadWrite.All Files.ReadWrite.All. offline_access is what makes Microsoft issue a refresh token. The .All in these delegated permissions means "every site the signed-in user can open", never more than the user's own SharePoint access. For a read-only setup, use offline_access User.Read Sites.Read.All Files.Read.All in the connector's OAuth settings (and add those two permissions to the app registration), authorize again, and disable sharepoint_create_list_item, sharepoint_update_list_item, sharepoint_create_list, sharepoint_create_folder, sharepoint_upload_text_file, sharepoint_rename_drive_item, sharepoint_move_drive_item.

Switched off at install: sharepoint_delete_list_item, sharepoint_delete_drive_item, sharepoint_create_sharing_link and sharepoint_invite. Sharing gives other people access to the organisation's files, and a document the AI reads could ask it to share something, so turn these on per tool only if the AI should share or delete. Deleted items go to the site's recycle bin.

Finding your way

1. Site: sharepoint_search_sites by keyword, sharepoint_get_site_by_path from an address (https://contoso.sharepoint.com/sites/Marketing → hostname contoso.sharepoint.com, site_path sites/Marketing), or sharepoint_get_root_site. A site id looks like contoso.sharepoint.com,<guid>,<guid>.
2. Files: sharepoint_list_drives lists a site's document libraries ("Documents" is the default one); then sharepoint_list_drive_root, sharepoint_list_folder, sharepoint_search_drive and sharepoint_get_drive_item with that drive id.
3. Lists: sharepoint_list_lists, then sharepoint_list_columns for the internal column names, then sharepoint_list_items.
4. Across everything: sharepoint_search uses Microsoft Search (KQL) over all sites, files and list items the user can see.

List items: values live in fields, keyed by the column's internal name, which can differ from the name shown (a column shown as "Due date" may be DueDate or Due_x0020_date). Filter with fields/<name>, e.g. fields/Status eq 'Open'. On lists over 5,000 items, filter and sort only on indexed columns; otherwise SharePoint refuses the query. Choice columns take one of their choices; lookup and person columns are written as <Name>LookupId with the id; dates as ISO 8601 in UTC.

Files: this connector does not download file content (Graph answers with a redirect to the whole file, whose size AnythingMCP cannot cap); give the user webUrl. sharepoint_upload_text_file creates text files up to about 4 MB, never overwriting with conflict_behavior fail (default). For cell values of Excel workbooks in a library, use the Microsoft Excel connector with the library's drive id.

Pages: sharepoint_list_pages and sharepoint_get_page read modern site pages; text web parts carry HTML in innerHtml. Classic pages, news posting and page editing are not covered.

Pagination: responses carry @odata.nextLink while more results exist. Call the same tool again with the same arguments and skip_token set to the $skiptoken value in that link. sharepoint_search pages with from and size instead and reports moreResultsAvailable.

Rate limits: SharePoint throttles per user and per app and answers 429 or 503 with Retry-After; wait and retry, and prefer filtered list queries over reading whole lists.

Common errors

• 401 InvalidAuthenticationToken: the authorization was revoked or the client secret expired. Paste a new secret if needed and click Authorize with Provider again.
• 403 accessDenied: a permission from step 4 is missing or was not consented, or the user has no access to that site, library or list.
• 404 itemNotFound: wrong site, drive, list or item id, or a site path that does not exist. Personal Microsoft accounts get 404 or 400 for every SharePoint call.
• 400 invalidRequest on list items: a column name in fields or filter is not an internal name, or a value does not fit the column type. 400 with "cannot be referenced in filter or orderby as it is not indexed": index the column in the list settings, or filter on another column.
• 409 nameAlreadyExists: an item with that name exists in the folder.

Out of scope here: downloading file content, binary uploads, site creation and site permissions, list and column deletion, page editing, term store and content types.