open-webSearch MCP Server

Content

open-webSearch gives Claude and other MCP clients no-key web search and public content retrieval. It supports multiple search engines, public page fetching, GitHub README fetching, CLI use, stdio MCP mode, and an optional local daemon for repeated local HTTP calls.

Use it when Claude needs a lightweight search/fetch tool for public web research without provisioning a commercial search API key, while still keeping site terms, rate limits, query privacy, and local daemon exposure visible.

Source Review

• https://github.com/Aas-ee/open-webSearch
• https://raw.githubusercontent.com/Aas-ee/open-webSearch/main/README.md
• https://registry.npmjs.org/open-websearch
• https://raw.githubusercontent.com/Aas-ee/open-webSearch/main/LICENSE
• https://raw.githubusercontent.com/Aas-ee/open-webSearch/main/package.json
• https://raw.githubusercontent.com/Aas-ee/open-webSearch/main/docs/http-api.md
• https://raw.githubusercontent.com/Aas-ee/open-webSearch/main/src/index.ts
• https://raw.githubusercontent.com/Aas-ee/open-webSearch/main/src/config.ts
• https://raw.githubusercontent.com/Aas-ee/open-webSearch/main/src/tools/setupTools.ts
• https://raw.githubusercontent.com/Aas-ee/open-webSearch/main/src/utils/urlSafety.ts

These sources were reviewed on 2026-06-06. Prefer the live repository, README, npm metadata, license, package manifest, local daemon API docs, server entrypoint, runtime configuration, tool setup, and URL safety helpers for current setup and behavior details.

Features

• Search across supported engines such as Bing, DuckDuckGo, Brave, Exa, Baidu, CSDN, Juejin, Startpage, and Sogou.
• Limit available engines with ALLOWED_SEARCH_ENGINES.
• Fetch generic public web pages and Markdown content.
• Fetch GitHub repository README content.
• Fetch supported article pages from CSDN and Juejin.
• Run as a local MCP stdio server, a one-shot CLI, or an explicit local daemon.
• Configure request-only, auto, or Playwright search mode for Bing workflows.
• Reuse local or remote Playwright/Chromium sessions when manually configured.
• Route requests through an approved HTTP proxy when needed.
• Reject private or local network fetch targets through URL safety checks.

Installation

Configure a stdio-only MCP setup first:

{
  "mcpServers": {
    "open-websearch": {
      "command": "npx",
      "args": ["-y", "open-websearch@latest"],
      "env": {
        "MODE": "stdio",
        "DEFAULT_SEARCH_ENGINE": "duckduckgo",
        "ALLOWED_SEARCH_ENGINES": "duckduckgo,startpage,bing",
        "NPM_CONFIG_IGNORE_SCRIPTS": "true"
      }
    }
  }
}

Claude Code users can add it with:

claude mcp add open-websearch --env MODE=stdio --env DEFAULT_SEARCH_ENGINE=duckduckgo -- npx -y open-websearch@latest

Start the local daemon separately only when you need a reusable local HTTP service for scripts, skills, or plugin integrations.

Use Cases

• Search the public web without provisioning a search API key.
• Compare results across multiple search engines.
• Fetch the readable content of a specific public page after search.
• Retrieve a GitHub README for quick repository context.
• Use a local daemon for repeated search/fetch calls from nearby tooling.
• Route live search traffic through an approved proxy in restricted networks.
• Use Playwright fallback for pages that need browser-rendered content.

Safety and Privacy

open-webSearch is a public web search and fetch tool, so results can be noisy, blocked, biased, or incomplete. Verify important claims against primary sources and respect each site's current terms before repeated fetching.

Keep stdio mode as the default for MCP clients. If you start the daemon, treat it as local-only tooling, keep CORS narrow when enabled, and avoid connecting it to browser sessions or proxies that expose sensitive cookies or network routes.