Bocha AI Web Search MCP Server

What is Bocha AI Web Search MCP Server?

An MCP (Model Context Protocol) server that provides web search functionality through Bocha AI, a legal and compliant AI search service operating in the People's Republic of China, specifically designed for searching Chinese internet content. It returns web content and related images in either human-readable Markdown or raw JSON format.

How to use Bocha AI Web Search MCP Server?

Clone the repository, copy .env.example to .env, configure your Bocha API key, then run npm install && npm run build. Configure the server in your MCP client (e.g., mcp.json) with the node command pointing to the built index.js and the BOCHA_API_KEY environment variable set accordingly.

Key features of Bocha AI Web Search MCP Server

• Web search for Chinese internet content with summaries and images
• Flexible time range filtering (OneDay, OneWeek, OneMonth, OneYear, noLimit)
• Optional detailed text summaries for search results
• Dual output: Markdown (human-readable) or JSON (structured data)
• Configurable result count (1–50, default 10)
• Robust error handling with clear messages

Use cases of Bocha AI Web Search MCP Server

• Retrieve current Chinese web content and news with configurable freshness
• Obtain structured search data for programmatic processing in applications
• Generate readable search summaries for end users in Markdown format
• Search Chinese internet sources while complying with PRC regulations

FAQ from Bocha AI Web Search MCP Server

What makes this server different from general web search tools?

It is specifically built for searching Chinese internet content and operates as a legal, compliant AI search service within the People's Republic of China, adhering to all relevant Chinese regulations and standards.

What are the runtime dependencies?

Node.js and npm must be installed. The server does not require a separate database or additional services beyond the Bocha AI API.

How do I provide authentication?

Set the BOCHA_API_KEY environment variable with your Bocha AI API key when starting the server or in the MCP client configuration.

Can I control the number of search results?

Yes, the count parameter accepts a number between 1 and 50, defaulting to 10.

What output formats are supported?

Two formats: Markdown (default, human-readable with links and image previews) and raw JSON (when raw_json is set to true for programmatic processing).