Open Deep Research

Use it when

• You need long-running web research on a topic synthesized into a full written report and manually collating dozens of search results is too slow
• You want an open source deep research agent you can run with your own choice of model provider, search API, or MCP servers instead of a closed vendor product

What it solves

Not the fit when

• Requires model credentials and usually a search API, native web-search provider, or configured MCP server; provider usage can incur charges
• Selected models must support structured output and tool calling
• Local Ollama and OpenRouter require separate setup beyond the default quickstart
• Search results and generated reports can contain incorrect, stale, or prompt-injected material; verify consequential claims against the underlying sources
• The documented roughly 20 to 100 USD estimate is for running all 100 Deep Research Bench examples, not a price guarantee for ordinary research runs
• quick single-fact lookup
• retrieval over a private document corpus
• offline research without model or search API access

Install

git clone https://github.com/langchain-ai/open_deep_research.git && cd open_deep_research && uv venv && source .venv/bin/activate && uv sync && cp .env.example .env

Invoke

Start the local LangGraph server with: uvx --refresh --from "langgraph-cli[inmem]" --with-editable . --python 3.11 langgraph dev --allow-blocking, then ask a question in the Studio UI messages field and click Submit

Alternatives

No reviewed alternatives recorded yet.