Higress AI-Search MCP Server

What is Higress AI-Search MCP Server?

Higress AI-Search MCP Server is a Model Context Protocol (MCP) server that provides an AI search tool to enhance AI model responses with real‑time search results. It integrates with the Higress ai‑search feature to search various engines and internal knowledge bases.

How to use Higress AI-Search MCP Server?

Install via uvx (automatic install from PyPI) or clone the repository and use uv locally. Configure environment variables: HIGRESS_URL (optional, default http://localhost:8080/v1/chat/completions), MODEL (required), and INTERNAL_KNOWLEDGE_BASES (optional). Add the server configuration to your MCP client (e.g., Cline or Claude Desktop) with the appropriate command and args.

Key features of Higress AI-Search MCP Server

• Internet search via Google, Bing, and Quark
• Academic search via Arxiv for scientific papers
• Internal knowledge base search
• Real‑time enhancement of AI model responses

Use cases of Higress AI-Search MCP Server

• Provide up‑to‑date web information to an AI assistant during a conversation
• Retrieve latest academic research papers on a given topic
• Answer queries using company‑internal knowledge bases
• Augment LLM responses with current data from multiple search sources

FAQ from Higress AI-Search MCP Server

What does the server do?

It provides an AI search tool that fetches results from external search engines and internal knowledge bases, then feeds them into an AI model to improve answer quality.

What are the prerequisites?

You need uv installed and a running Higress instance configured with both the ai‑search and ai‑proxy plugins.

Is it required to have a Higress service running?

Yes. The server connects to a Higress service at the URL specified by HIGRESS_URL (default http://localhost:8080/v1/chat/completions).

What search engines are supported?

Google, Bing, Quark (general web search) and Arxiv (academic search). Internal knowledge base search is also supported.

How do I configure the server?

Set environment variables HIGRESS_URL, MODEL (required), and optionally INTERNAL_KNOWLEDGE_BASES. Then add the server configuration to your MCP client.