ufn

The UniFuncs MCP Server exposes a suite of API tools for real-time web search, web content extraction, and deep research tasks. It includes web-search for on-demand internet queries with pagination and output formats, web-reader for extracting and formatting page content (with optional images and link summaries), and both synchronous and asynchronous deep search and deep research capabilities. You can run the MCP server via NPX, which downloads and runs the server package of UniFuncs, and provide your API key to authorize usage. The server exposes endpoints that accept structured input for each tool (e.g., query objects for web-search, URL and options for web-reader, and task-based flows for deep-search and deep-research). The asynchronous flows return a task_id that you poll to retrieve results, enabling long-running operations without blocking.

To use, start the server with your API key and, if you need streaming/SSE transport, enable the SSE endpoint as described in the setup. Once running, you can issue commands according to the tool reference section: perform a web search with a query, request a web page reader output, or create and poll async tasks for deep search and deep research. The server will respond with results in the requested format (json, markdown, or text) depending on your input for each tool.

Prerequisites:

• Node.js (and npm) installed on your system
• Internet access to fetch the MCP server package

Installation steps:

1. Install and run the MCP server via NPX with your UniFuncs API key:

npx -y @unifuncs/ufn-mcp-server

2. Alternatively, provide the API key through environment variables when starting the server:

UNIFUNCS_API_KEY=your_api_key npx -y @unifuncs/ufn-mcp-server

3. If you need SSE transport, set the environment variables or use the --sse flag as described in the README:

export UNIFUNCS_SSE_SERVER=true
export UNIFUNCS_SSE_SERVER_PORT=5656

npx @unifuncs/ufn-mcp-server --sse

4. Verify the server is running by checking logs and connecting to the API endpoints specified in the documentation (e.g., web-search, web-reader, deep-search, deep-research).

5. Optionally, switch to a specific packaging method by providing credentials and environment variables as needed by your deployment environment.

Tips and notes:

• Always provide your UNIFUNCS_API_KEY when initiating the MCP server to authorize requests.
• If using SSE, ensure the environment variables UNIFUNCS_SSE_SERVER and UNIFUNCS_SSE_SERVER_PORT are set as needed.
• The deep-search and deep-research tools are task-based. For long queries, use the asynchronous endpoints (create_task and query_task) to avoid blocking.
• The web-search tool supports freshness filters (Day/Week/Month/Year) and pagination (1-50 results per page). Choose the output format (json, markdown, text) as needed for downstream processing.
• Web-reader can include images, provide link summaries, and output Markdown-formatted content for easy inclusion in docs.
• For production deployments, consider setting a stable hosting environment and rate-limiting to manage API usage and billing.
• If you encounter issues, consult the official documentation and GitHub repository linked in the README for troubleshooting and updates.