Favicon of Open Deep Research

Open Deep Research

Self-hosted AI research agent that supports Ollama, cloud models and MCP tools. MIT licensed; archived and no longer maintained.

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 The project is archived and no longer maintained.

The Python agent runs on a local LangGraph server, with a browser interface through LangGraph Studio. Its MIT license lets you use and modify the open source code. Hosted deployment is also available through LangGraph Platform, and Open Agent Platform provides an interface for people who want to configure agents without writing code. You can host that interface yourself or use the public demo with your own API keys.

Model choice extends to local models through Ollama, cloud providers such as OpenAI and Anthropic, and OpenRouter. Models must support tool calling and structured outputs. You can choose separate models to summarize search results, carry out research, condense findings and write the final report.

For web research, the agent uses Tavily by default and also supports OpenAI and Anthropic's native web search. MCP support connects it to additional tools and servers. Running the agent locally doesn't make those external services local: cloud model calls and online searches still go to the selected providers.

The project includes evaluation tooling for Deep Research Bench through LangSmith, so developers can compare generated reports against expert reference reports.

Similar to Open Deep Research