Yandex Webmaster MCP

What Yandex Webmaster MCP MCP server does

Yandex Webmaster MCP MCP server exposes Yandex Webmaster information and selected account actions to AI applications through MCP. It is intended for sites managed in Yandex Webmaster and focuses on organic search data rather than advertising, keyword-demand research, or general traffic analytics.

The available functions cover several parts of site management:

• Listing accessible sites and viewing their IKS, indexed pages, excluded pages, and issue counts.
• Reviewing search queries with impressions, clicks, and average positions over selected dates.
• Comparing search visibility between mobile and desktop devices.
• Inspecting crawl HTTP errors, important-page indexing states, sitemap status, and recrawl queues.
• Viewing examples of external pages that link to a site.
• Adding sites or sitemaps, beginning ownership verification, and submitting URLs for recrawling.

How it works

The Yandex Webmaster MCP MCP server starts by working with the sites available to the authenticated account. Each site has a host_id; the server can infer it from a request or use YANDEX_WEBMASTER_HOST_ID when that variable is set.

Authentication can be initiated from the conversation. The assistant directs the user to Yandex for sign-in, receives a one-time code that remains valid for 10 minutes, and checks which sites are accessible. Statistics and diagnostics require confirmed ownership of the relevant site. The server only returns data for sites the token can access.

Most operations are read-only. Actions that alter Webmaster state include adding a site, starting ownership verification, adding a sitemap, and placing a page in the recrawl queue. The queue operation consumes the site's daily quota. A raw_request capability can call API paths without dedicated tools; depending on the HTTP method, that request can also modify data or delete a site or sitemap. Clients may ask for confirmation, but confirmation behavior depends on the client.

Setup and configuration

Node.js 20 or newer is required. The package can be started with npx, which downloads it on the first run, so a separate package installation is not needed. A typical MCP configuration uses the command npx with arguments -y mcp-yandex-webmaster@latest.

For Claude Desktop, add a yandex-webmaster entry to claude_desktop_config.json under mcpServers, then restart the application. Cursor accepts the same command and arguments in either the user-level ~/.cursor/mcp.json file or a project-level .cursor/mcp.json file.

After adding the server, ask the client to connect Yandex Webmaster and complete the sign-in flow. For automated setups, a previously available token can be configured according to the project's technical documentation. YANDEX_WEBMASTER_HOST_ID is optional and can select a default site context.

Tools and capabilities

The Yandex Webmaster MCP MCP server provides 20 tools spanning sites, diagnostics, search queries, indexing, sitemaps, external links, and account connection. Natural-language requests can retrieve recent query performance, identify serious diagnostic issues, show indexing changes, inspect sitemap reading errors, and report remaining recrawl capacity.

Use it for questions such as which pages are indexed, which queries generated impressions during a period, what crawl errors Yandex encountered, or whether a sitemap is registered. Use write-capable operations carefully: ownership verification requires placing the code supplied by Webmaster, and deletion through a raw API request may be irreversible.

The service does not provide Yandex Metrica traffic data, Wordstat search-demand data, or advertising-cabinet information. Statuses such as HOST_NOT_VERIFIED, HOST_NOT_LOADED, and HOST_NOT_INDEXED indicate access, loading, or indexing conditions rather than zero-valued metrics.