Content
# MCP Server Product Name: Bocha
## Version
v1
## Product Description
### Short Description
Bocha is a search engine for AI, allowing your AI applications to obtain high-quality world knowledge from nearly 10 billion web pages and ecological content sources, covering various fields such as weather, news, encyclopedia, medical, train tickets, and images.
### Long Description
Bocha is a search engine for AI, allowing your AI applications to obtain high-quality world knowledge from nearly 10 billion web pages and ecological content sources, covering various fields such as weather, news, encyclopedia, medical, train tickets, and images.
## Category
Web Search
## Tags
Search, News, Weather, Encyclopedia
## Tools
### Tool1: Bocha Web Search
#### Detailed Description
Search the entire network for information and web links from Bocha Search, and return results including web page titles, web page URLs, web page summaries, website names, website icons, publication times, image links, etc.
#### Parameters Required for Debugging
Input:
- query: Search term (required)
- freshness: Search for web pages within a specified time range (optional values YYYY-MM-DD, YYYY-MM-DD..YYYY-MM-DD, noLimit, oneYear, oneMonth, oneWeek, oneDay. The default is noLimit)
- count: Number of results returned (1-50, default is 10)
Output:
- Web page title, web page link, web page summary, publication time, website name
### Tool2: Bocha AI Search
#### Detailed Description
Based on Bocha Web Search, AI identifies the semantics of search terms and additionally returns structured modal cards of vertical domain content, such as weather cards, calendar cards, encyclopedia cards, and dozens of other modal cards, which are better in terms of semantic recognition, timeliness of search results, and content richness.
#### Parameters Required for Debugging
Input:
- query: Search term (required)
- freshness: Search for web pages within a specified time range (optional values YYYY-MM-DD, YYYY-MM-DD..YYYY-MM-DD, noLimit, oneYear, oneMonth, oneWeek, oneDay. The default is noLimit)
- count: Number of results returned (1-50, default is 10)
Output:
- Web page title, web page link, web page summary, publication time, website name, modal card
## Adaptable Platforms
Fangzhou, python, Claude, Cursor, etc.
## Service Activation Link
You need to go to [Bocha AI Open Platform](https://open.bochaai.com) and log in to get the API KEY.
## Authentication Method
API Key
## Installation and Deployment
### Step 1: Download the Code to Local
```bash
git clone git@github.com:BochaAI/bocha-search-mcp.git
```
### Step 2: Configure in the Client
#### Claude Desktop
On MacOS: `~/Library/Application\ Support/Claude/claude_desktop_config.json`
On Windows: `%APPDATA%/Claude/claude_desktop_config.json`
```json
"mcpServers": {
"bocha-search-mcp": {
"command": "uv",
"args": [
"--directory",
"/path/to/bocha-search-mcp",
"run",
"bocha-search-mcp"
],
"env": {
"BOCHA_API_KEY": "sk-****"
}
}
}
```
### Step 3: Use in the Client
### Step 4: Debug Local Service (Optional)
```bash
npx @modelcontextprotocol/inspector uv --directory /path/to/bocha-search-mcp run bocha-search-mcp
```
## Customer Cases
Currently, Bocha has served a total of **3000+ enterprise users** and **20000+ developer users**, and has become the **official online search provider of DeepSeek** and the **officially recommended search API of Alibaba, Tencent, and ByteDance**. Currently, it **undertakes more than 60% of the online search requests of AI applications in China**.
Bocha's search content sources include nearly 10 billion web pages on the entire network, as well as ecological cooperation content (including short videos, news, encyclopedia, weather, medical, train tickets, hotels, restaurants, attractions, enterprises, academics, etc.). Bocha will continue to carry out co-creation cooperation with various platforms in content ecology, intelligent body creation, etc., to provide rich and colorful answers to the search questions of Bocha users.
## Common Problems
### What capabilities can the Bocha Web Search API service provide?
Bocha Web Search provides general search capabilities for the entire network. You can search the entire network for information and web links from Bocha Search, and return results including web page titles, web page URLs, web page summaries, website names, website icons, publication times, image links, etc. The maximum number of web pages returned for each search result is 50 (count 50).
Traditional search engines use a keyword + bidding ranking mechanism search algorithm. The goal of the search results is not to directly provide users with the correct answers, but to attract users to click to obtain advertising revenue.
Bocha is a new generation search engine based on multi-modal hybrid search and semantic ranking technology. It supports natural language search methods for AI application scenarios, and the goal of the search results is to provide clean, accurate, and high-quality answers.
Bocha's semantic ranking technology is based on the Transformer architecture and will rank search results according to their semantic relevance to user questions. Since the large model is also a Transformer architecture, it judges the semantic relevance of the context to the user's question and makes trade-offs, so the large model ultimately prefers the search results provided by Bocha.
Currently, Bocha's search effect is the closest search engine to Bing Search API in China. Since Bing Search API data will go overseas (no domestic Region), is expensive (15 US dollars/thousand times) and does not provide text summaries (only 50-100 word snippets), many domestic enterprise customers have switched from Bing to Bocha.
### What capabilities can the Bocha AI Search API service provide?
Based on Bocha Web Search, Bocha AI Search AI identifies the semantics of search terms and additionally returns structured modal cards of vertical domain content, such as weather cards, calendar cards, encyclopedia cards, and dozens of other modal cards, which are better in terms of semantic recognition, timeliness of search results, and content richness.
The currently supported modal card types include: weather, encyclopedia, medical, perpetual calendar, train, constellation zodiac, precious metals, exchange rate, oil price, mobile phone, stock, automobile, etc.
Taking stock information as an example, real-time stock data is generally not available on web pages, and structured modal cards are needed to support it. The Bocha AI Search API can additionally output structured data modal cards of stock prices on the basis of providing web page information. Through the structured data provided by the modal cards, the accuracy of users' answers to timeliness questions in AI applications can be further enhanced.
