What is Local Deep Research?
Local Deep Research is an AI research assistant that searches the web, academic databases and private document collections to answer complex questions. It offers quick summaries, detailed research and structured reports with citations. Its agentic research strategy lets the model select search engines and decide when to synthesize findings.
The application runs through Docker, Docker Compose or a Python package and provides a browser-based interface. It can use locally running models through Ollama or connect to cloud providers. Retrieved papers and web pages can be downloaded into a library, indexed and searched alongside other documents. A REST API, Python access and an MCP server support use outside the web interface.
Each user has an isolated SQLCipher database, with encryption keys derived from the account password. Local deployment does not automatically keep all traffic private: cloud models can receive prompts, and public search engines receive queries even through a self-hosted search proxy. Egress controls provide best-effort restrictions rather than guaranteed network isolation.
Key capabilities
- Research strategies Supports quick summaries, detailed analysis and autonomous research with dynamic search engine selection.
- Multi-source search Queries academic services such as arXiv and PubMed, web search engines and private document collections.
- Document library Downloads research sources, extracts text and indexes documents for semantic retrieval in later research.
- Model selection Connects to local inference through Ollama and to supported cloud language model providers.
- Report delivery Produces cited reports and supports exports including PDF and Markdown.
- Programmatic access Exposes research through a REST API, Python interfaces and an MCP server compatible with Claude Desktop.
Where Local Deep Research fits
Local Deep Research fits literature exploration, technical investigations and research across internal document collections. Its library allows retrieved material to become searchable input for subsequent work, while the MCP server makes research available from compatible AI assistants.
Before deployment, operators should verify model hardware requirements, search engine access, credentials and the intended egress scope. Sensitive-document workflows require checking where inference and embeddings actually run, including whether local endpoints relay requests to hosted services. The unencrypted development fallback does not verify account passwords and is unsuitable for shared deployments. PDF export on Windows requires an additional Pango setup.
Install
pip install local-deep-researchReviews of Local Deep Research
Be the first to share your experience with Local Deep Research.
Write a review
Loading review form…