Yandex Disk MCP Server
A Yandex Disk MCP server built for finding large files and cleaning up via the API, bypassing local sync
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- Can permanently delete files and empty the trash with no way to undo it
- Can move and rename files on Yandex Disk
- The token grants full disk access within the OAuth application's scopes
Install
Manual install
npm install
npm run buildInstall dependencies and build src into dist/index.js.
This is third-party code. Review the repository files before installing.
What it does
The server gives an agent direct access to Yandex Disk via the REST API rather than a synced Finder folder, so it can see files that are not on the local machine. It has eight tools: disk quota, folder listing sorted by size, a recursive tree scan for the largest files and folders, downloading and uploading a file, deleting to trash or permanently, moving and renaming, and emptying the trash, which also counts against quota. The README explicitly notes what it cannot do: toggle the offline/cloud-only status of an already-synced file, since that is an internal feature of the desktop client, not part of the public API.
Who it is for. For people who want an agent to find what is taking up space on Yandex Disk and clean it up without the web interface.
Good fit when
- You need to find the largest files and folders on Yandex Disk, including ones not synced locally
- You need to free up space, including emptying the trash
- You need to download or upload a specific file directly to the cloud without waiting for local sync
Not a fit when
- You need to toggle the cloud-only status of an already-synced file: the server cannot do this by API design
- You need a guaranteed full scan on very large folder trees: the scan can be partial when it hits the max_items_to_scan limit
Example request
Find the ten largest files on Yandex Disk and suggest what could be deletedLimitations
No license is declared in the repository. There is no npm package; install and build are from source via npm install and npm run build. Yandex Disk has no recursive folder-size method, so the tree scan is paginated and can be partial on very large trees, flagged by capped in the response. Permanent deletion and emptying the trash are irreversible.
How to disable. Remove the server with claude mcp remove yandex-disk and revoke the Yandex Disk token at yandex.ru/dev/disk/poligon.
MCP
- Transport
- stdio
- Authentication
- API key
| Environment variables | |
|---|---|
| YANDEX_DISK_TOKEN required, secret | Yandex Disk OAuth token, obtained at yandex.ru/dev/disk/poligon |
Security check
- Can permanently delete files and empty the trash with no way to undo it
- Can move and rename files on Yandex Disk
- The token grants full disk access within the OAuth application's scopes
README in short
The Russian README explains the server's purpose: direct access to the Yandex Disk API bypassing local sync, for analyzing used space, targeted uploads and downloads, and managing structure. It gives a table of eight tools, a section on what the server cannot do, instructions for getting an OAuth token at yandex.ru/dev/disk/poligon with a warning not to confuse it with the Client ID and Client Secret, npm install steps and Claude Code registration with user scope. It explains the recursive tree-scan behavior and the max_items_to_scan limit.
FAQ
Can a file be switched to cloud-only mode?
Not directly, that is an internal feature of the Yandex Disk desktop client, not part of the REST API. A similar result comes from uploading the file straight to the cloud with the upload tool.
Why is a scan of large folders sometimes incomplete?
Yandex Disk has no recursive folder-size method, so the server walks the tree itself page by page, and on very large trees the max_items_to_scan limit can cut the scan short, flagged as capped.
Related
A self-hosted knowledge base with block-level references and a built-in MCP server for connecting AI agents to your notes
A CLI for every Google Workspace API with JSON output and agent skills: Drive, Gmail, Calendar, Sheets and more
Local search over Markdown notes, docs and meeting transcripts: keywords, semantic search and reranking, with an MCP server
A task manager for AI-driven development: breaks a PRD into dependent tasks and guides the agent through them via MCP or CLI