home / mcp / higress ai-search mcp server

Higress AI-Search MCP Server

An MCP server enhances AI responses with real-time search results via Higress ai-search.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "cr7258-higress-ai-search-mcp-server": {
      "command": "uvx",
      "args": [
        "higress-ai-search-mcp-server"
      ],
      "env": {
        "MODEL": "qwen-turbo",
        "HIGRESS_URL": "http://localhost:8080/v1/chat/completions",
        "INTERNAL_KNOWLEDGE_BASES": "Employee handbook, company policies, internal process documents"
      }
    }
  }
}

You can run the Higress AI-Search MCP Server to enable real-time internet, academic, and internal knowledge searches as part of your model’s responses. This server augments AI outputs with up-to-date web results and internal knowledge bases, making your AI interactions more informed and context-aware.

How to use

You deploy the Higress AI-Search MCP Server as an MCP endpoint and connect it with an MCP client. The server supports multiple search sources, including general internet search, academic search for papers, and internal knowledge bases. When you ask questions, the MCP client passes prompts to the server which then integrates relevant search results into the model’s responses. You can run it in a development environment or in production, using the provided startup options.

How to install

Prerequisites you need before installation:

Install uvx or uv to run the MCP server locally.

Configure Higress with ai-search and ai-proxy plugins.

Choose one of the startup options below and place the corresponding MCP configuration in your environment.

Option 1: Using uvx

{
  "mcpServers": {
    "higress_ai_search_mcp_server": {
      "command": "uvx",
      "args": [
        "higress-ai-search-mcp-server"
      ],
      "env": {
        "HIGRESS_URL": "http://localhost:8080/v1/chat/completions",
        "MODEL": "qwen-turbo",
        "INTERNAL_KNOWLEDGE_BASES": "Employee handbook, company policies, internal process documents"
      }
    }
  }
}

Option 2: Using uv with local development

{
  "mcpServers": {
    "higress_ai_search_mcp_server": {
      "command": "uv",
      "args": [
        "--directory",
        "path/to/src/higress-ai-search-mcp-server",
        "run",
        "higress-ai-search-mcp-server"
      ],
      "env": {
        "HIGRESS_URL": "http://localhost:8080/v1/chat/completions",
        "MODEL": "qwen-turbo",
        "INTERNAL_KNOWLEDGE_BASES": "Employee handbook, company policies, internal process documents"
      }
    }
  }
}

Configuration and usage notes

The MCP server exposes a straightforward interface for your MCP client. You can continue to adjust the model, knowledge bases, and Higress service URL through environment variables as needed. If you need to change the base Higress URL or the model used for response generation, update the corresponding environment variables in the startup configuration and restart the MCP server.

Available tools

internet_search

Search the web using Google, Bing, or Quark to fetch general information for responses.

academic_search

Query Arxiv and related academic databases to gather papers and scientific references.

internal_knowledge_search

Search internal knowledge bases to retrieve organization-specific information.