---

name: web-search description: > Perform DuckDuckGo web searches and return structured results. WHEN: User wants to "search the web", "find information online", "look up", "research topic". WHEN NOT: Know exact URL (use browser_navigate), crawling sites (use scrape_url). version: 0.1.0

web_search - DuckDuckGo Web Search

Core Concept

mcp__plugin_kg_kodegen__web_search performs web searches using DuckDuckGo and returns structured results with titles, URLs, and snippets. Uses headless browser with stealth configuration to avoid bot detection.

Key Parameters

Usage Example

Copy{ "query": "rust async programming tutorial" }

Response Format

Copy{
  "success": true,
  "query": "rust async programming tutorial",
  "results_count": 10,
  "results": [
    {
      "rank": 1,
      "title": "Async Programming in Rust - The Rust Book",
      "url": "https://doc.rust-lang.org/book/ch16-00-concurrency.html",
      "snippet": "Learn about async/await syntax and futures in Rust..."
    },
    {
      "rank": 2,
      "title": "Tokio Tutorial",
      "url": "https://tokio.rs/tokio/tutorial",
      "snippet": "Build reliable network applications with Tokio..."
    }
  ]
}

Result Fields

Performance Notes

• First search: ~5-6 seconds (browser launch)
• Subsequent searches: ~3-4 seconds
• Max results: 10 per query
• Uses DuckDuckGo to avoid CAPTCHA issues

Use Cases

1. Research topics - Find documentation and tutorials
2. Discover libraries - Search for packages and tools
3. Find solutions - Look up error messages and fixes
4. Gather information - Research before code generation

Workflow Examples

Research Before Coding

Copy{ "query": "rust error handling best practices" }

Find Library Documentation

Copy{ "query": "serde json serialization rust" }

Debug Error Messages

Copy{ "query": "rust borrow checker error E0382" }

Remember

• Returns up to 10 results per search
• Results include title, URL, and snippet
• First search is slower (browser initialization)
• Uses DuckDuckGo - good for avoiding CAPTCHAs
• For known URLs, use browser tools instead
• For site-specific search, use scrape_url with SEARCH action