serpdive● live

**Real-time web search built for AI agents.** One tool, `serpdive_search`, returns the actual content of the best pages — extracted, cleaned, and sized for LLMs — not a list of links to crawl. ## Why SERPdive On a [public, replayable 1,000-question benchmark](https://github.com/edendalexis/serpdive-benchmark) against Tavily: - **Same speed** (median ~1.7 s server-side) - **20.2% fewer tokens** fed to your LLM per search - 🏆 **60.7% of decided quality duels** won Same questions, same judge, replayable end to end on your machine. ## The tool `serpdive_search(query, model?, answer?, max_results?)` - **`query`** — the search, in any language; localization is automatic - **`model`** — `mako` (default): the fact-carrying sentences of each page, fast. `moby`: the full readable text, for deep research - **`answer`** — `true` adds a direct answer synthesized from the sources (cited on `moby`) - **`max_results`** — hard cap on results, 1 to 10 Every result ships as `{ url, title, date?, content }` — quote and cite straight from the response. Failed searches are never billed. ## Get started Free API key, no card required → [serpdive.com/dashboard/keys](https://serpdive.com/dashboard/keys) - Hosted server: `https://mcp.serpdive.com` (this listing) - Local stdio: `npx -y serpdive-mcp` ([GitHub](https://github.com/serpdive/serpdive-mcp)) - Docs: [serpdive.com/docs](https://serpdive.com/docs)

listed onSmithery