You can run a high-availability MCP server that aggregates multiple search engines, handles intelligent failover, and provides a unified API for querying diverse sources. This server is designed for concurrency and robust content processing, making it suitable for applications needing reliable, multi-engine search results with consistent formatting.

How to use

You interact with the Search Fusion MCP Server through the MCP client by configuring one or more MCP endpoints. Start the server locally or connect to a remote MCP endpoint, then issue search requests and content fetch operations. The server automatically selects engines based on availability and priority, performs failover when needed, and returns results in a unified JSON structure. Use the available tools to search, fetch pages, query Wikipedia, or access archived pages via Wayback, all backed by concurrency safeguards and real-time engine status.

How to install

pip install search-fusion-mcp

Or install from source

git clone https://github.com/sailaoda/search-fusion-mcp.git cd search-fusion-mcp pip install -e .

## Configuration & MCP integration

The server supports environment variable based MCP configuration. You can run one or more MCP endpoints as separate servers. The following MCP configurations illustrate how to start the server in different modes and pass necessary API keys via environment variables.

{ "mcp": { "mcpServers": { "search_fusion": { "command": "search-fusion-mcp", "env": { "GOOGLE_API_KEY": "your_google_api_key", "GOOGLE_CSE_ID": "your_google_cse_id", "SERPER_API_KEY": "your_serper_api_key", "JINA_API_KEY": "your_jina_api_key", "EXA_API_KEY": "your_exa_api_key", "BING_API_KEY": "your_bing_api_key", "BAIDU_API_KEY": "your_baidu_api_key", "BAIDU_SECRET_KEY": "your_baidu_secret_key" } } } } }

This example shows a local MCP endpoint started via a command and configured with environment variables for each supported engine. You can also configure the server using a Python invocation if you install from source, as shown in the next snippet.

Security and best practices

Respect API rate limits for each engine, enable sensible timeouts, and monitor real-time engine status to detect failures early. The system includes automatic failover, semaphore-based concurrency control, and per-request timeouts to prevent resource exhaustion. Always secure your API keys and avoid exposing them in client-side code.

Tools and endpoints

You have access to specialized tools for searching and content retrieval. Use the functions described to perform web searches, fetch and paginate content, retrieve available engines, search Wikipedia, and query archived pages.

Troubleshooting

If you encounter high latency or timeouts, check engine availability and adjust the semaphore limit or per-engine timeouts. Review the real-time engine status to identify which engines are active, and consider reordering engines based on current performance. Ensure all required API keys are set correctly in the MCP environment.

Available tools

search

Perform web searches with intelligent engine selection and failover.

fetch_url

Fetch and process web content with intelligent pagination and multi-method fallback.

get_available_engines

Get current status and availability of all search engines.

search_wikipedia

Search Wikipedia articles for entities, people, places, concepts, etc.

search_archived_webpage

Search archived versions of websites using Wayback Machine.