Skip to main content
Tavily Chat searches the web and answers a question about when to visit Japan, with citations and source links.

Try Tavily Chat

Tavily Chat

Features

  • Context-aware search: Ask follow-up questions using context from recent messages.
  • Intelligent parameter selection: The agent chooses search queries and relevant primary-source domains from your question.
  • Content snippets: Tavily returns excerpts from web pages as context for the agent’s answer.
  • Parallel search: Primary-source and independent-source searches run at the same time.
  • Source attribution: Open the web pages behind an answer using numbered citations and source links.
  • Streaming: See search queries as they run and read the answer as it is generated.

How it works

Tavily Chat pairs an AI agent with Tavily Search. The agent turns your question and recent conversation into two search queries: one for primary sources and one for independent coverage.
A question flows from the React interface through query planning, two parallel Tavily searches, source selection, and answer generation before returning to the chat.
The Node.js backend runs both searches in parallel, removes duplicate URLs, and selects excerpts from multiple publishers. The agent uses these excerpts to write an answer with citations. The backend streams search updates, sources, and answer text to the React interface.