Yandex Search

This MCP server lets you access Yandex Search through a lightweight, real-time interface so your AI assistants can browse the web and pull up-to-date information safely. It provides two practical tools for search and retrieval that you can integrate with your preferred MCP client.

How to use

Connect to the Yandex Search MCP server from your MCP client to perform live web searches and generate AI-assisted results. You can use two primary tools: ai_search_post for real-time searches with AI-generated synthesis, and web_search_post for real-time searches with explicit sources. Use the HTTP endpoint for remote access or run a local MCP server and connect via the provided tools. Ensure you have the correct API key and folder ID to authorize your requests.

How to install

Prerequisites you need before installing:

• Node.js 16+ for MCP development and remote integration
• Python 3.10+ for alternative local setup
• Docker 20+ or Podman 3+ if you plan to run the server in containers
• Access to a Yandex Search API key and a valid folder ID for authentication

Option 1. Use Yandex Search MCP with Docker or Podman

1. Build the container docker build -t yandex-mcp-server-image:latest .- or -podman build -t yandex-mcp-server-image:latest .
2. Run with your credentials (replace placeholders with your values) via an MCP client configuration that references the remote image. The MCP client will expose the endpoints and tools once the container is running.

Option 2. Use Python + MCP Config
1. Install required libraries

pip install -r requirements.txt

2. Create an MCP config referencing the Python server script and your credentials. Example shown below is a direct MCP config using a stdio endpoint.

Option 3. Run the local MCP server directly (no container) after you have set your API key and folder ID as environment variables:

python3 server.py```
The server will listen for incoming MCP requests on stdin and respond on stdout.

## Configuration and examples

Use these explicit configuration samples to connect from your MCP client. They illustrate both HTTP (remote) access and a local stdio-based startup using npx.

{ "mcpServers": { "yandex_search_http": { "type": "http", "url": "https://d5dj4o5pbnqgca1d546v.cmxivbes.apigw.yandexcloud.net:3000/sse", "args": [] } } }

{ "mcpServers": { "yandex_search_stdio": { "type": "stdio", "command": "npx", "args": [ "-y", "mcp-remote", "https://d5dj4o5pbnqgca1d546v.cmxivbes.apigw.yandexcloud.net:3000/sse", "--header", "ApiKey:<your_api_key>", "--header", "FolderId:<your_folder_id>" ] } } }

## Security and credentials

Keep your API key and folder ID secure. Do not share them in logs or public configuration files. Use environment variables or secret management where possible, and apply the principle of least privilege to the service account used for the MCP connection.

## Troubleshooting

If the server fails to start, verify that all prerequisites are installed and that you have valid credentials. Double-check network access to the Yandex API endpoint and ensure your API key has the required scope (yc.search-api.execute) and that the folder’s service account has search-api.editor rights.

## Notes

Two tools are exposed through the MCP: ai\_search\_post for AI-generated answers based on search results and web\_search\_post for standard web search with sources. Use these tools in your MCP client by selecting the appropriate tool and issuing a search query.

## Available tools

### ai\_search\_post

Performs a real-time web search and returns an AI-generated answer based on the search results using Yandex Yazeka AI model.

### web\_search\_post

Performs a real-time web search and returns an answer with sources.