Local Deep Researcher is an open-source research assistant from the LangChain team that runs entirely on your own machine: it searches the web, reads the sources and writes a cited report, all driven by a local language model. Published under the MIT license, it has earned more than 9,300 stars on GitHub and was actively developed through 2026.
Hosted deep-research products send your questions โ and the model doing the work โ to someone else’s cloud. This project shows the same workflow fully local: built on LangGraph, it takes a topic, plans search queries, fetches and summarizes web results in a loop that reflects on gaps and searches again, then assembles a structured markdown report with sources. The language model runs locally through Ollama or another local server, so the reasoning stays on your hardware while only the web searches themselves leave the machine.
Local Deep Researcher is free and open source under the MIT license. Some optional search providers require your own API key, and local models run on your own hardware.
The assistant is deliberately small and hackable: its research loop is a compact LangGraph graph, so developers routinely fork it to add private data sources, change the report format or plug in a stronger local model. It also doubles as a teaching example for how agentic search-and-reflect loops are structured in production systems.
It is for privacy-conscious researchers, analysts and developers who want deep-research-style reports without handing their questions and documents to a hosted service.