Web Search
Use the built-in web_search and web_fetch tools for web research.
Searching
Use web_search with a query string:
- •
query: Search terms (required) - •
count: Number of results (1-10, default 5) - •
country: 2-letter country code for localized results (e.g., "US", "DE") - •
freshness: Filter by time ("pd" = past day, "pw" = past week, "pm" = past month)
Returns titles, URLs, and snippets.
Fetching Pages
Use web_fetch to extract content from a URL:
- •
url: The page URL (required) - •
extractMode: "markdown" (default) or "text" - •
maxChars: Truncate long pages
Returns readable content without HTML clutter.
Workflow
- •Search for relevant pages
- •Pick the most promising URLs from results
- •Fetch and read the content
- •Synthesize information for the user
Tips
- •Use specific search terms for better results
- •Fetch multiple sources for verification
- •Summarize rather than dumping raw content
- •Cite sources when providing information