Xmlriver MCP

xmlriver-mcp

MCP server for XMLRiver — Google/Yandex SERP parsing and Yandex Wordstat keyword frequency data via XML API.

mcp-name: io.github.artgas1/xmlriver-mcp

What it does

Gives Claude / Cursor / Windsurf direct access to:

• Google SERP parsing (organic, ads, FAQ, knowledge graph, AI Overview) for any country / region / device
• Yandex SERP parsing (Russian-speaking markets — primary use case)
• Yandex Wordstat keyword frequency, history, similar queries (Yandex's keyword volume tool)
• Indexing check — is this URL in Google/Yandex index?
• Account ops — balance, tariff, cost per 1k requests

First MCP for XMLRiver — fills a gap for Russian SEO research and Yandex-aware analysis. Pay-as-you-go (~25 ₽ / 1000 requests on Basic tariff).

Demo

Claude queries wordstat_query and parses real frequency data from Yandex. Same flow works for google_search, yandex_search, indexing checks, and account ops.

Quickstart

uvx xmlriver-mcp

Configuration

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{

"mcpServers": { "xmlriver": { "command": "uvx", "args": ["xmlriver-mcp"], "env": { "XMLRIVER_USER": " ", "XMLRIVER_KEY": " " } } } }

Claude Code

Add to project .mcp.json:

{

"mcpServers": { "xmlriver": { "command": "uvx", "args": ["xmlriver-mcp"], "env": { "XMLRIVER_USER": " ", "XMLRIVER_KEY": " " } } } }

Cursor

Edit ~/.cursor/mcp.json (global) or .cursor/mcp.json (project):

{

"mcpServers": { "xmlriver": { "command": "uvx", "args": ["xmlriver-mcp"], "env": { "XMLRIVER_USER": " ", "XMLRIVER_KEY": " " } } } }

Tools

All tools are read-only (annotated with readOnlyHint: true). No destructive operations.

⚠️ `google_maps_search` is unverified against a live response. Every setab=maps request answered code 500 while it was written (2026-08-04), including a deliberately invalid one that should have returned code 108 — plain Google and Yandex searches worked in the same minutes, so the mode appears unavailable service-side. The request shape and response parsing follow the documentation and are covered by unit tests; both still need a live check.

Deliberately not wrapped: raw=page (returns the full HTML page — bypasses parsing and returns megabytes into the model's context) and the deferred-response mode (delayed / req_id) — an MCP call is synchronous, so a request id the caller must poll for has nowhere useful to go.

Authentication

1. Register at https://xmlriver.com
2. Top up balance (minimum ~100 ₽ to start)
3. Get your user (numeric ID) and key (40-char hex) from the dashboard
4. Set XMLRIVER_USER and XMLRIVER_KEY env vars in your MCP client config

Security note: XMLRiver API is HTTP-only (not HTTPS). The key is rotatable from the dashboard if compromised.

Pricing context

Use get_balance and get_cost to monitor spend before bulk operations.

Common use cases

• SEO position tracking — yandex_search(query="...", region=213) for own/competitor ranking
• Keyword research — wordstat_query(query="купить iphone", history_period="monthly") for demand validation + seasonality
• Featured snippet hunting — google_search(additional_blocks="faqsnippet,knowledge_graph,zeroposition") to see what owns the answer box
• Long-tail discovery — search_suggestions(phrases=[...]) with a prefix sweep (seed + each letter of the alphabet) pulls phrasings Wordstat never reports
• Indexation monitoring — check_url_indexed(url="https://your-site.com/new-page") after publishing
• Cross-region comparison — same query, different region/country for Yandex/Google to see geographic variance

Local development

git clone https://github.com/artgas1/xmlriver-mcp

cd xmlriver-mcp uv sync --all-extras

Run unit tests (no API key needed)

uv run pytest tests/unit -v

Run integration tests (requires XMLRIVER_USER / XMLRIVER_KEY)

XMLRIVER_USER=... XMLRIVER_KEY=... uv run pytest tests/integration -v -m integration

MCP Inspector — interactive

XMLRIVER_USER=... XMLRIVER_KEY=... npx @modelcontextprotocol/inspector uv run python -m xmlriver_mcp.server

MCP Inspector — CLI smoke test (list tools)

npx @modelcontextprotocol/inspector --cli "uv run python -m xmlriver_mcp.server" --method tools/list

Architecture

• Stack: Python 3.10+ / FastMCP / httpx / tenacity / pydantic
• Transport: stdio (default)
• No external SDK dependency — direct REST via httpx + custom XML parser
• Retry strategy: 3 attempts with exponential backoff on network errors (not on HTTP 4xx)
• Logging: stderr only (stdio protocol requires stdout for JSON-RPC)

License

MIT

Contributing

PRs welcome. Open an issue first for substantial changes.

Acknowledgements

• XMLRiver — for the underlying API
• Anthropic MCP — for the protocol