Skip to content

MindSearch

Build your own multi-agent web search engine

Visit official site

What MindSearch is

MindSearch is described as an open-source multi-agent web search engine framework built on large language models, with performance similar to Perplexity.ai Pro. The framework decomposes a query into atomic sub-questions in a graph and expands the graph from WebSearcher results.

What you can do with it

  • Deploy your own search engine with GPT, Claude or InternLM2.5-7b-chat
  • Browse hundreds of web pages for deep knowledge discovery
  • Show a detailed solution path for the final response
  • Choose among React, Gradio, Streamlit and Terminal interfaces

Application scenarios

  • A developer builds a personalised AI search engine
  • An enterprise raises the search efficiency of an internal knowledge base
  • A research institution runs deep academic material searches

Getting started

  • Install the Python libraries listed in requirements.txt
  • Set up the API with a FastAPI server and configure language and model formats
  • Pick a front end from React, Gradio, Streamlit or Terminal
  • Start the back end service and check that the API responds

Pros & cons

✓ What we like

  • Several interfaces including React, Gradio, Streamlit and Terminal
  • Works with both closed-source and open-source LLMs

! What to watch out for

  • Deployment requires installing Python libraries from requirements.txt
  • A FastAPI server must be configured for the API

FAQ

Which language models can it use?

Closed-source models such as GPT and Claude, or open-source models such as InternLM2.5-7b-chat.

What interfaces are offered?

React, Gradio, Streamlit and Terminal are listed as UI options.

Last reviewed: 2026-11-15

More AI search engine tools

View all →

How we review