Local Deep Researcher is a web research and report generation tool that runs entirely locally using language models hosted through Ollama or LMStudio.
The tool addresses the need for autonomous research workflows without relying on cloud-based APIs or proprietary services. It works by iteratively generating web search queries, gathering results, summarizing findings, identifying knowledge gaps through reflection, and repeating this cycle a configurable number of times. The final output is a markdown report with cited sources. The system supports multiple search backends including DuckDuckGo (which requires no API key), SearXNG, Tavily, and Perplexity, and can use either JSON mode or tool calling depending on the model's capabilities.
Developers should choose this tool if they need research automation without external dependencies or API costs, and if they have the hardware to run a local language model. It suits projects requiring privacy-preserving research workflows or environments where cloud connectivity is restricted. The tool works with any model available through Ollama or LMStudio, giving flexibility in model selection. Unlike cloud-based research assistants, it keeps all processing local and allows users to select their preferred search provider.
The project shows active development with recent additions of tool calling support and compatibility with new model families. The codebase includes configuration management through environment variables and integrates with LangGraph for workflow orchestration. Documentation covers multiple setup paths for different local model platforms and search tool options.
