Valyu MCP Js

🚀 Valyu MCP Server

A Model Context Protocol (MCP) server that provides access to knowledge retrieval and feedback APIs. This server enables large language models to search proprietary and web sources for information and submit feedback on transactions.

🚀 Quick Start

The Valyu MCP Server offers APIs for knowledge retrieval and feedback submission, allowing large - language models to access information from various sources and manage transaction feedback.

✨ Features

Available Tools

• knowledge: Search proprietary and/or web sources for information.

  • Required Parameters:
    • query (string): The question or topic to search for.
    • search_type (string): The type of search source ("proprietary", "web", or "all").
    • max_price (number): The maximum allowed price per thousand queries (CPM).
  • Optional Parameters:
    • data_sources (array of strings): A list of index names to search.
    • max_num_results (integer): The number of results to return after re - ranking.
    • similarity_threshold (number): The minimum similarity score to include results.
    • query_rewrite (boolean): Whether to rewrite the query for better performance.

• feedback: Submit user feedback on a transaction.

  • Required Parameters:
    • tx_id (string): The transaction ID for which to provide feedback.
    • feedback (string): The user's feedback text.
    • sentiment (string): The sentiment rating ("very good", "good", "bad", "very bad").

📦 Installation

Using Docker

docker pull ghcr.io/tiovikram/valyu-mcp-server
docker run -i --rm -e VALYU_API_KEY=your-api-key ghcr.io/tiovikram/valyu-mcp-server

📚 Documentation

Configuration

Environment Variables

• VALYU_API_KEY (required): Your Valyu API key.

Configuration on Claude.app

Add the following to your Claude settings:

{
  "name": "knowledge",
  "arguments": {
    "query": "What is quantum computing?",
    "search_type": "all",
    "max_price": 0.5,
    "data_sources": ["valyu/valyu-arxiv", "valyu/valyu-wikipedia"],
    "max_num_results": 5
  }
}

{
  "name": "feedback",
  "arguments": {
    "tx_id": "12345abcdef",
    "feedback": "The information was very helpful and accurate.",
    "sentiment": "very good"
  }
}

npx @modelcontextprotocol/inspector node dist/index.js