In short. Kagi Research's Librarian retrieves and analyzes documents or URLs in depth within a multi-step workflow that can expose tool calls and references.
Definition: Kagi Research's Librarian retrieves and analyzes information from documents or URLs in depth. Kagi places it among the specialized agents used in a multi-step research process.
Key facts
- Kagi's Research documentation lists Librarian as the tool that retrieves and analyzes information from documents or URLs in depth.
- The same documentation describes approximately 15 agent roles, including research agents, final-response agents and librarian agents for in-depth source analysis.
- Kagi Research can perform up to 5 steps per query, with up to approximately 15 tool calls, so it can chain operations around source analysis.
- Kagi's November 2025 announcement says the research workflow identifies searches, runs multiple searches at the same time and synthesizes the findings into answers.
- The announcement says research-assistant answers are expected to be sourced and referenced, with attribution of citation relevance in the final answer.
How does Librarian handle a source URL?
The documentation gives Librarian a specific retrieval and analysis role: it works with information from documents or URLs in depth. A URL therefore supplies an address for source material that the tool can retrieve and analyze as part of a Kagi Research request. The page does not specify a URL length limit, a maximum number of URL inputs or the fetching rules applied to a particular website.
The same tool list assigns file reading and analysis to File Processing, including text, PDF, image, video and audio formats. That separation matters when a request includes a document file alongside a source URL: the documentation names Librarian for the URL or document information and File Processing for the listed file-format operations.
The source address shown here is the documentation page that names the URL capability:
https://help.kagi.com/kagi/ai/kagi-research.html
What happens around the retrieved source?
Kagi describes Research and Quick as meta models built from specialized sub-models. Research breaks a request into multiple steps and uses different tools and models. Its available-tool list includes Kagi Search, Librarian, a Python Interpreter, a reasoning model, Wolfram|Alpha, image tools, Maps Search and File Processing.
Kagi's official announcement describes the wider research workflow as identifying what to search for, executing simultaneous searches in different languages when needed, and synthesizing the findings. The announcement also says the research assistant can call tools during the answer process. Librarian is therefore one part of a workflow that can combine URL analysis with search, reasoning or other operations.
How can you inspect the sources used?
Kagi says the tool calls appear in purple dropdown boxes in the research interface. Opening those boxes exposes the search results used during the process. The same announcement says answers are expected to be sourced and referenced and that Kagi adds attribution of citation relevance to the final answer.
This disclosure describes how to inspect the research trail after a response: open the tool call, review the returned search results and then follow the references in the answer. Kagi's announcement frames that design as a way to encourage people to look further into the answer and accelerate their research process.
How does Librarian differ between Quick and Research?
Kagi's comparison table includes Librarian, labelled "in-depth source analysis", for both Kagi Quick and Kagi Research. It lists average research and analysis time as under 5 seconds for Quick and 30 seconds or more for Research. The table gives Research additional response planning, research planning, Python, image, image-editing and Wolfram|Alpha capabilities.
Related terms
FAQ
Can Librarian analyze a URL and a file in the same request?
Kagi's documentation assigns Librarian to information from documents or URLs and separately lists File Processing for text, PDF, image, video and audio formats. It does not describe a special combined-input syntax.
Is Librarian available in Kagi Quick?
Yes. Kagi's comparison table marks Librarian for both Kagi Quick and Kagi Research, while Research adds planning and other tools.
Does Kagi publish the URL-fetching limits for Librarian?
The documentation identifies documents and URLs as Librarian inputs. It leaves URL size, input-count and site-fetching limits unspecified.
Sources
- Kagi, "Kagi Research Assistants", undated documentation, accessed 17 October 2026
- Kagi, "Introducing Kagi Assistants", 20 November 2025
Check how a page reads to answer engines. Open the GEO score analyzer