Local Deep Research
Self-hosted AI research assistant for web, papers and private documents. Runs on Windows, macOS and Linux with Ollama or cloud models. MIT licensed.
Local Deep Research is a self-hosted AI research assistant for people who need cited answers drawn from academic papers, the web and their own documents. It can produce a quick summary or pursue a complex question through repeated searches, then assemble a structured report. It's open source under MIT.
Its LangGraph research mode lets the model choose search queries and sources as the investigation develops. Academic search includes arXiv, PubMed and Semantic Scholar; other sources include SearXNG, Wikipedia and GitHub. You can save downloaded papers and pages in an encrypted library and search them alongside live web results in later research.
The browser interface runs on Windows, macOS and Linux, with Docker also supported. Model backends include Ollama, llama.cpp and OpenAI-compatible endpoints, with cloud providers available too. CPU-only use is supported, and Docker supports NVIDIA GPUs on Linux. ARM64 is supported; x86-64 machines need an AVX-capable processor.
Each user has a separate encrypted SQLCipher database that also holds their API keys. The app has no telemetry or tracking. Local models keep inference on your hardware, while web searches send queries to the selected search services and cloud models send requests to your chosen provider.
Follow-up chat retains research context across turns. You can revisit saved research, export reports as PDF or Markdown, and schedule topic digests. LangChain retrievers connect existing knowledge bases, while a local MCP server gives Claude Desktop and Claude Code access to its research tools.
