Open Deep Research

Open Deep Research is a self-hosted AI agent that searches for information and writes research reports for teams that want to choose their models and tools. The project is archived and no longer maintained.

• Model choice: Use local models through Ollama or providers such as OpenAI, Anthropic and OpenRouter, with separate model choices for research, summarization and report writing.
• Web research: Search through Tavily by default or use OpenAI and Anthropic's native web search, and connect additional research tools through MCP.
• Browser interfaces: Work with the agent through LangGraph Studio or make it available through a self-hosted Open Agent Platform interface.
• Report evaluation: Compare generated reports with expert reference reports using Deep Research Bench evaluation tooling through LangSmith.

The Python agent runs on a local LangGraph server and is open source under MIT. Selected models must support tool calling and structured outputs. Local operation doesn't make web research offline: searches and cloud model calls go to the selected external providers.

Hosted deployment is available through LangGraph Platform. The public Open Agent Platform demo uses your own API keys.