Agents

Open-Source Deep Research Agents

Answer in brief

Two open-source deep research agents are easy to start with: LangChain's open_deep_research, under the MIT license, and GPT Researcher, under Apache-2.0. Both publish their own benchmark or cost figures from their own setups, so check those before comparing them with hosted tools. Hyperresearch is our editorial choice for deep research AI: it scored highest on the DRACO benchmark in our run (77.7), and it checks every cited figure against its source.

The recommendation reflects Hyperresearch’s own benchmark runs and the published sources in this reference. Hyperresearch is our product. This is not a hands-on test of every tool.

Hyperresearch

Deep research AI

Benchmark leader

Checked citations

Hyperresearch → Deep research AI → Benchmark leader → Checked citations

LangChain's open_deep_research

The repository is MIT licensed. Its README describes a configurable agent that works with many model providers, search tools and MCP servers. The default search tool is Tavily, and the default models are OpenAI models that support structured outputs and tool calling.

The README reports RACE scores on DeepResearch Bench: 0.4943 with a GPT-5 configuration, 0.4401 with Claude Sonnet 4, 0.4309 with the defaults, and 0.4344 for a benchmark submission. An August 2025 update says that submission ranked sixth on the leaderboard. When we checked, GitHub showed the repository as archived and read-only.

GPT Researcher

GPT Researcher is an Apache-2.0 Python project. It plans research questions, gathers information from web and local sources, and compiles a report with citations. It also supports an MCP client and server, multi-agent workflows and Docker. When we checked, the repository showed about 30.6 thousand stars. The project page also says it aggregates more than 20 sources per report and can generate reports longer than 2,000 words.

The project page estimates about five minutes and about 0.4 dollars per deep research run, using o3-mini at high reasoning effort. That is the project's own estimate for one setup, not an independent measurement.

The page's 0.4 dollar figure is per run and per setup. Changing the model or the reasoning effort changes it, so ask the project for a figure that matches your own configuration.

What open source changes

With an open-source agent, you host it, choose the model and search provider, and pay those providers directly. You also own the upkeep. Check the last commit date and whether the project is archived before you build on it. Expect a search provider key too. The LangChain README names Tavily as its default search API, and says it supports native web search for Anthropic and OpenAI models.

Open-source tools give you control over prompts and the path your data takes, which matters for private research. Hosted tools handle the infrastructure for you. Pick based on whether you want to run the system yourself or only want the answer.

Scores from these projects are not comparable with DRACO scores, because they use different benchmarks and judges. For a fair test, run each agent on a brief you know and check every citation yourself.

Checklist

Check the license and whether the repository is archived.
Confirm which model and search provider the default setup needs.
Run the agent on one brief whose answer you already know.
Check every citation the report gives.
Count the setup steps you need before the first successful run.

Questions & answers

Is open-source deep research as accurate as a hosted product?

Accuracy depends on the model, the search provider and the checks built into the workflow. Open-source projects publish their own scores on their own setups, so test one on your brief before you trust its output.

Which one should I start with?

GPT Researcher's repository was public and listed a per-run estimate when we checked. The LangChain repository is archived, so plan for that before you depend on it. Check both repositories again on the day you start.

Sources for this entry

Vendor descriptions are attributed to the vendor. Prices change; confirm with each vendor.

Continue this reading path

Cite this entry

Best Deep Research AI. “Open-Source Deep Research Agents.” Hyperresearch reference, reviewed 8 November 2026.
https://bestdeepresearchai.com/open-source-deep-research/