MCP Directory

Web Search MCP Server

Locally hosted multi-engine web search and page content extraction for local LLMs — no API keys required.

Unverified
stdio (local)
No auth
Stale
TypeScript
View repo 980

Add to your client

Copy the config for your MCP client and paste it into its config file.

Install / run
npm install && npx playwright install && npm run build

Paste into ~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "web-search-mcp-server": {
      "command": "node",
      "args": [
        "/path/to/extracted/web-search-mcp/dist/index.js"
      ]
    }
  }
}

Step-by-step guides: Add to Claude Desktop · Add to Cursor · Add to Windsurf

Before you start

  • Node.js 18.0.0 or higher
  • npm 8.0.0 or higher
  • Playwright browsers installed via 'npx playwright install'

About Web Search MCP Server

A simple, locally hosted web search MCP server built for local LLMs. It performs multi-engine web search (Bing, then Brave, then DuckDuckGo) without requiring any API keys, and can fetch and extract full page content from search results. A smart request strategy switches between Playwright browser instances and fast axios HTTP requests to maximise reliability, with concurrent multi-page extraction and automatic browser cleanup. Behaviour is configurable through environment variables for content length, timeouts, browser count, browser types, fallback thresholds, and search-quality/relevance checking.

Tools & capabilities (3)

full-web-search

Main tool. Takes a query and an optional number of results (1-10, default 5), performs a multi-engine web search (Bing, then Brave, then DuckDuckGo), fetches full page content from each result URL with concurrent processing, and returns structured data with search results and extracted content. Includes HTTP/2 error recovery and reduced timeouts.

get-web-search-summaries

Lightweight alternative. Takes a query and an optional number of results (1-10, default 5), performs the same optimised multi-engine search, and returns only the search result snippets/descriptions without following links to extract full page content.

get-single-web-page-content

Utility tool. Takes a single URL (and optional maxContentLength), follows it, and extracts the main page content while removing navigation, ads, and other non-content elements. Useful for getting detailed content from a known webpage.

What this server can do

Web Search MCP Server provides tools for these capabilities — tap one to see every MCP server that does the same:

Search the web

When to use it

  • Giving a local LLM (e.g. in LM Studio or LibreChat) live web search without paying for a search API
  • Comprehensive research that needs full page content extracted from multiple results at once
  • Quick lookups where only search result snippets are needed for speed
  • Extracting clean main content from a specific known webpage URL

Security notes

Runs locally with no API keys. Performs browser automation and HTTP requests to third-party search engines and arbitrary result URLs; some networks block browser automation. Content length is capped via MAX_CONTENT_LENGTH (default 500000 characters).

Web Search MCP Server FAQ

Do I need any API keys?

No. The server uses direct connections to search engines and web pages (browser automation via Playwright plus axios HTTP requests), so no API keys are required.

Which MCP clients does it support?

It has been developed and tested with LM Studio and LibreChat. The README notes it has not been tested with other MCP clients.

Which models work best?

Recent models designated for tool use work best; Qwen3 and Gemma 3 give the best results. Llama 3.2, recent Llama 3.1, and recent Deepseek R1 (e.g. 0528) also work, while older Llama and Deepseek R1 versions may not.

What if content length causes odd behaviour?

Set the MAX_CONTENT_LENGTH environment variable (e.g. "10000") in your mcp.json. Other env vars like DEFAULT_TIMEOUT, MAX_BROWSERS, BROWSER_TYPES, and BROWSER_FALLBACK_THRESHOLD let you tune timeouts, memory, and reliability.

How do I install it?

Download the latest release zip from the Releases page, extract it, then run 'npm install', 'npx playwright install', and 'npm run build' in the extracted folder. Point your mcp.json command to the built dist/index.js file.

#web-search#local-llm#lmstudio#librechat#playwright#content-extraction#bing#brave#duckduckgo#typescript

Alternatives to Web Search MCP Server

Compare all alternatives →
Fetch (Reference)
Search & Web
87k

Official MCP reference server that fetches a URL and returns its content as clean Markdown, with chunking.

Verified
stdio (local)
No auth
Python
2 tools
Updated 20 days agoRepo
Firecrawl
Search & Web
6.6k

Official Firecrawl MCP server — scrape, crawl, map, search, and structured extraction for any LLM client.

Verified
stdio (local)
API key
TypeScript
8 tools
Updated 18 days agoRepo
Exa Search
Search & Web
4.6k

Exa's neural web search and crawling MCP server — runs locally via npx or as a hosted remote endpoint.

Verified
stdio (local)
API key
TypeScript
6 tools
Updated 19 days agoRepo

Compare Web Search MCP Server with:

vs Fetch (Reference)vs Firecrawlvs Exa Searchvs ArXiv MCP Server