Self-Hosted Deep Research Studio

Productivity

Ask a question and AI agents search the web and write a research report with cited sources; runs on your own machine with Docker Compose, bound to 127.0.0.1. A bring-your-own-key product: you supply your own OpenAI and Tavily keys. Derived from a frozen snapshot of an Apache-2.0 open-source project.

Subscribe & download See how to start

SubscriptionUS$49 per month

Last run2026-11-12

In practice

Readiness3 / 5

What you need

Few changes needed3 / 5

Who it is for

Individuals, researchers and small teams who want AI agents to gather web material on their own machine and turn it into a cited report, and who already have or are willing to get OpenAI and Tavily keys.

How to start

$ sh scripts/quickstart/init-env.sh
$ docker compose up -d --build

Browse the files

Files in the product repository at the pinned commit. A few house files are left out of the zip.

File overview929 paths

▸backend31

Dockerfile
Procfile
__init__.py
chat/
memory/
report_type/
requirements.txt
run_server.py
runtime.txt
server/
…

▸deep_agents73

BENCHMARK.md
README.md
SHARE_SUMMARY.md
__init__.py
agent.py
benchmark.py
benchmark_data/
benchmark_results/
breadth_benchmark.py
drb_generate.py
…

▸docs129

CNAME
README.md
babel.config.js
blog/
discord-bot/
docs/
docusaurus.config.mjs
npm/
package.json
pydoc-markdown.yml
…

▸evals52

README.md
__init__.py
context_filter/
hallucination_eval/
quality_eval/
simple_evals/

▸frontend222

README.md
index.html
nextjs/
pdf_styles.css
scripts.js
static/
styles.css
vendor/

▸gpt_researcher132

__init__.py
actions/
agent.py
config/
context/
document/
llm_provider/
mcp/
memory/
observability.py
…

▸mcp-server1

README.md

▸multi_agents41

README.md
__init__.py
ag2/
agent.py
agents/
langgraph.json
main.py
memory/
package-lock.json
package.json
…

▸scripts1

quickstart/

▸skills1

gpt-researcher/

▸terraform12

ecr-setup/
github-actions-setup/
main.tf
outputs.tf
variables.tf
versions.tf

▸tests182

__init__.py
backend/
chat/
conftest.py
docs/
documents-report-source.py
gptr-logs-handler.py
report-types.py
research_test.py
retrievers/
…

▫ATTRIBUTION.md

▫CODE_OF_CONDUCT.md

▫CONTRIBUTING.md

▫CURSOR_RULES.md

▫Dockerfile

▫Dockerfile.fullstack

▫LICENSE

▫PRODUCT.md

▫Procfile

▫QUICKSTART.md

▫README-ja_JP.md

▫README-ko_KR.md

▫README-ru_RU.md

▫README-zh_CN.md

▫README.md

▫SECURITY.md

▫SECURITY.upstream.md

▫citation.cff

▫cli.py

▫delivery-manifest.txt

▫docker-compose.yml

▫json_schema_generator.py

▫langgraph.json

▫main.py

▫poetry.toml

▫pyproject.toml

▫render.yaml

▫requirements.txt

▫Config and hidden files 24

Latest passing run

Clone → Checkout → Toolchain → Run studio

Pre-purchase Disclosure

What you are buying

A tool where you ask a question and AI agents search the web and write a research report with cited sources: a frontend and a backend service, brought up with Docker Compose on your own machine.

Derived from an open-source project under the Apache-2.0 licence, and not the original project itself.

Delivery is a frozen snapshot: one zip, taken from the product repository at revision 40683e1.

Known limitations

The two services are bound to the local machine only over HTTP (web on port 3000, backend on port 8000); TLS and a reverse proxy are yours to bring for any wider access
The research features need your own OpenAI and Tavily keys; without them research tasks cannot run, although the web home page still opens
Research output and logs are stored in the unzipped folder, which has to stay in place while the services run
The first build needs internet access to download Python and Node dependencies
Backend dependencies resolve to the latest compatible versions at build time, so builds made at different times may differ
Dependency licences were not scanned package by package
Pinned to the delivered snapshot; it does not follow upstream.

The liability cap, the scope of what is promised and the statutory floor are governed by the Terms of Service.

Vetting record

Everything below was checked against the pinned commit ddd4db3. Every verdict carries its full reasoning, an evidence link and the checker; failures are published, not hidden — do not take our word for it, verify it yourself.

1 PASS4 WARN0 FAIL

Source ReviewWARN

Read: QUICKSTART.md, docker-compose.yml, .env.example, scripts/quickstart/init-env.sh, and the key lines of the Dockerfiles. The Python backend and frontend source were not read file by file.

Script SafetyWARN

The generator script was read in full: it copies .env.example into .env, filling values from your shell if exported. The Dockerfiles were only checked on key lines (pinned base images, a non-root user for the backend); no download-and-pipe-to-shell pattern was looked for beyond that. Other upstream scripts were not read.

RecencyWARN

This is a frozen snapshot that does not follow upstream. The date of the pinned upstream commit and recent activity were not checked.

LicensePASS

SPDX:Apache-2.0

The product repository has a LICENSE file, and PRODUCT.md records the upstream license as Apache-2.0. The file text was not compared line by line against the upstream copy.

Dependency LicensesWARN

Dependencies are declared in the requirements files and the frontend package files and fetched at build time. They were not scanned for licenses.