mcp-onedrive-sharepoint
Health Uyari
- No license — Repository has no license file
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Low visibility — Only 5 GitHub stars
Code Uyari
- network request — Outbound network request in package.json
- process.env — Environment variable access in src/auth/setup-auth.ts
- process.env — Environment variable access in src/config/index.ts
- network request — Outbound network request in src/graph/client.ts
Permissions Gecti
- Permissions — No dangerous permissions requested
Bu listing icin henuz AI raporu yok.
MCP server + CLI for OneDrive and SharePoint via Microsoft Graph (device-code auth, 32 tools, ods CLI)
MCP OneDrive/SharePoint Server
MCP server for Microsoft Graph focused on OneDrive, SharePoint and related document workflows.
This repository now has working onboarding and quality commands on a clean clone:
npm run buildnpm run lintnpm testnpm run cinpm run setup-auth
What is implemented
The server exposes 32 MCP tools grouped into:
- Files:
list_files,download_file,upload_file,create_folder,move_item,delete_item,search_files,get_file_metadata,share_item,copy_item - SharePoint:
discover_sites,list_site_lists,get_list_schema,list_items,get_list_item,create_list_item,update_list_item,delete_list_item - Utilities:
health_check,get_user_profile,list_drives,global_search,batch_operations - Advanced:
advanced_share,manage_permissions,check_user_access,sync_folder,batch_file_operations,storage_analytics,version_management,excel_operations,excel_analysis
Recent foundation improvements
This version includes real structural improvements instead of documentation-only changes:
- fixed
npm run setup-authto call the real TypeScript auth setup flow - added working ESLint configuration
- added executable tests with Node's built-in test runner
- aligned README and
.env.examplewith the real environment variables and scripts - introduced reusable Graph helpers for:
- consistent MCP JSON/error envelopes
- pagination extraction from Microsoft Graph responses
- resolving OneDrive/SharePoint resources by
driveId,siteId,itemIdand path
- updated key listing/search flows to return pagination metadata and support
driveId
Requirements
- Node.js 18+
- A Microsoft Entra ID / Azure app registration for delegated device-code authentication
Installation
git clone <repository-url>
cd mcp-onedrive-sharepoint
npm install
cp .env.example .env
Configuration
The server reads the following environment variables:
MICROSOFT_GRAPH_CLIENT_ID=your_app_client_id
MICROSOFT_GRAPH_TENANT_ID=common
MICROSOFT_GRAPH_SCOPES=Files.ReadWrite.All,Sites.ReadWrite.All,Directory.Read.All,User.Read,offline_access
MICROSOFT_GRAPH_BASE_URL=https://graph.microsoft.com/v1.0
MICROSOFT_GRAPH_TIMEOUT=30000
MICROSOFT_GRAPH_MAX_RETRIES=3
MICROSOFT_GRAPH_CACHE_ENABLED=true
MICROSOFT_GRAPH_CACHE_TTL=3600
Notes:
- use
MICROSOFT_GRAPH_TENANT_ID=commonfor multi-tenant/device-code onboarding - use a specific tenant id if you want tenant-scoped sign-in
- delegated scopes are what the current auth flow uses
Authentication setup
Run:
npm run setup-auth
The script:
- reads
MICROSOFT_GRAPH_CLIENT_ID/MICROSOFT_GRAPH_TENANT_IDfrom.envwhen present - prompts for missing values
- starts Microsoft device-code login
- stores the token through the existing auth layer
Development commands
npm run build
npm run lint
npm test
npm run ci
npm start
npm run ci is the local verification entrypoint and is also what GitHub Actions runs on every PR/push.
MCP behavior notes
Root site inclusion
discover_sites.includePersonalSite=true currently attempts to append the tenant root SharePoint site (/sites/root) when it is available to the authenticated user.
It does not discover or synthesize a personal OneDrive site.
Pagination
The following tools now expose consistent pagination metadata in their JSON payloads:
list_filessearch_filesdiscover_siteslist_site_listslist_items
When Microsoft Graph returns @odata.nextLink, the response includes:
pagination.returnedpagination.limitpagination.totalCountwhen availablepagination.nextPageTokenpagination.hasMore
Pass pageToken back to the same tool to continue paging.
Drive/site targeting
Core file listing/search/download flows now accept:
siteIdfor a SharePoint site's default drivedriveIdfor a specific document library or drive- path-based addressing where supported
This is the current foundation for moving beyond a /me/drive-only model.
Example tool inputs
List files from a specific drive
{
"driveId": "b!abc123",
"path": "/Shared Documents",
"limit": 50
}
Continue a paginated file listing
{
"driveId": "b!abc123",
"pageToken": "https://graph.microsoft.com/v1.0/drives/b!abc123/root/children?$skiptoken=..."
}
Search files in a site drive
{
"siteId": "contoso.sharepoint.com,123,456",
"query": "quarterly report",
"limit": 25
}
List SharePoint list items with pagination
{
"siteId": "contoso.sharepoint.com,123,456",
"listId": "9c6b8b70-0000-0000-0000-111111111111",
"orderBy": "Created desc",
"limit": 100
}
Quality status
Validated locally after these changes:
npm run build✅npm run lint✅npm test✅
Current limitations
- authentication still depends on real Microsoft Graph credentials and an interactive device-code login
- only the most critical listing/search foundations were migrated to the new pagination/resource helpers in this pass; other tools still use older direct endpoint construction
- advanced tools exist, but this PR focuses on onboarding, contract clarity and foundation hardening rather than a full architectural rewrite
Yorumlar (0)
Yorum birakmak icin giris yap.
Yorum birakSonuc bulunamadi