Kagi for content research
Content research starts with gathering sources and reading them closely, and Kagi brings an ad-free, privacy-respecting search index to that work. It is our fifth pick of five here, which is honest placement: the tooling is narrow, two tools, but the index quality is the reason it earns a spot at all.
Where a writer values a clean, commercial-incentive-free set of results over breadth of features, Kagi is a reasonable backend. For most content research workflows, though, the servers ahead of it do more of the synthesis and crawling work an agent needs to turn sources into a draft.
How Kagi fits
Kagi exposes two tools. kagi_search_fetch runs web, news, video, podcast, and image search with domain and date filters and Kagi lenses, and it can return page extracts inline, so an agent can search and read in one call. kagi_extract pulls the full content of a known URL back as clean markdown when you already have the link.
The limit is scope. Kagi searches and extracts; it does not synthesize an answer with citations, and it does not crawl a site systematically. Perplexity is stronger when you want a cited synthesis rather than raw results. Exa fits better for retrieval tuned to how an agent consumes results. Firecrawl is the pick when research means crawling whole sites into markdown at volume, and Tavily is built around search results shaped for LLM pipelines. Choose Kagi when result quality and a clean index are the priority and you are fine doing the synthesis yourself.
Tools you would use
| Tool | What it does |
|---|---|
| kagi_search_fetch | Runs web, news, video, podcast, and image search with optional page extracts, domain and date filters, and Kagi lenses. |
| kagi_extract | Fetches the full content of a known URL and returns it as clean markdown. |
FAQ
- Does Kagi's MCP server write the research summary for me?
- No. Its two tools, kagi_search_fetch and kagi_extract, find and return source material as clean markdown. Turning those sources into a synthesized, cited summary is the agent's job. If you want the server itself to return a cited answer, Perplexity is the closer fit.
- When is Kagi the right research backend over the other picks?
- When result quality and a privacy-respecting, ad-free index matter more than features. For systematic site crawling Firecrawl wins, for cited synthesis Perplexity wins, and for LLM-shaped retrieval Exa and Tavily fit better. Kagi is the clean-index choice.