Aas-ee/open-webSearch
Free一个基于多引擎搜索结果的模型上下文协议(MCP)服务器，支持免费网络搜索，无需API密钥。
FreeFree tier
Inputs: url
About Aas-ee/open-webSearch
open-websearch is an open-source toolkit that provides an MCP server, CLI, and local daemon for live web search and content retrieval without requiring any API keys. It aggregates results from multiple search engines including Bing, Baidu, DuckDuckGo, Exa, Brave, Juejin, StartPage, and Sogou, with configurable number of results and customizable default engine. It also supports fetching individual article content from CSDN, GitHub README files, and generic HTTP(S) pages, returning structured results with titles, URLs, and descriptions. The tool includes HTTP proxy configuration for accessing restricted resources and can be used as a skill-guided agent workflow for autonomous web interaction.
Key Features
Multi-engine web search (Bing, Baidu, DuckDuckGo, Exa, Brave, Juejin, StartPage, Sogou)
No API keys or authentication required
MCP server for integration with Claude Desktop, Cherry Studio, Cursor, and other MCP clients
CLI for one-shot local commands and shell scripts
Local daemon as a long-lived HTTP service with health and search endpoints
Skill-guided agent workflows for automated setup and usage
HTTP proxy configuration for accessing restricted resources
Configurable number of results per search and customizable default search engine
Fetch individual article content from CSDN, GitHub README, and generic HTTP(S) pages
Returns structured results with titles, URLs, and descriptions
Pros & Cons
Pros
- Completely free with no API key requirements
- Supports a wide range of search engines including both mainstream and niche options
- Flexible deployment as MCP server, CLI, or local daemon
- Skill-guided workflows simplify agent integration
- HTTP proxy support for restricted network environments
- Structured and configurable search results
Cons
- Some search engines (e.g., linux.do) are temporarily unsupported
- Requires Node.js and npm for installation (not explicitly stated but implied for npm-based tool)
Best For
Connecting to AI desktop clients like Claude Desktop, Cherry Studio, or Cursor via MCPRunning one-shot web search commands in terminal scriptsDeploying a long-lived local HTTP service for agent-based search and content retrievalEquipping AI agents with skill-guided web search workflows
FAQ
Does open-websearch require API keys or authentication?
No, open-websearch works without any API keys or authentication. It uses free public web search engines.
What search engines does open-websearch support?
It currently supports Bing, Baidu, DuckDuckGo, Exa, Brave, Juejin, StartPage, and Sogou. linux.do is temporarily unsupported.
How can I use open-websearch with AI clients like Claude Desktop?
You can use the MCP server mode to connect open-websearch to any MCP-compatible client such as Claude Desktop, Cherry Studio, or Cursor.
Can open-websearch fetch content from specific websites?
Yes, it can fetch article content from CSDN, GitHub README files, and generic HTTP(S) pages, returning Markdown or structured content.