Compose red team infrastructure and Active Directory ranges on a canvas, then compile to a complete Terraform and Ansible working directory you run yourself. Validate TTPs and tooling at production-ops level. GCP, AWS, and Azure.

A visual canvas for red team infrastructure and cyber ranges. Compose a topology, export a complete, runnable working directory of Terraform and Ansible, and deploy it from your own machine. redStackPRO never holds your cloud credentials.
redStackPRO is a validation range for TTPs and tooling at production-operations level, not a training or CTF product.
[!IMPORTANT] redStackPRO is in prerelease (beta). Features are still moving before 1.0. GCP, AWS, and Azure are tested end to end; Proxmox and ESXi are on the roadmap. Expect rough edges, and pin to a released version if you need stability.
Hit a rough edge, or have feedback? Open an issue at Issues. If it was a deploy, attach the scrubbed
logs/deploy-*.logthe run wrote (it records versions, provider, and where it stopped, with secrets removed) so it can be analyzed fast. Your reports shape the release.
redStackPRO puts attack infrastructure and target ranges on the same canvas. The two
canvas modes are Offense (attack infrastructure) and Defense (defensive AD ranges); the
export names its handoff OFFENSE-BRIEFING.md or DEFENSE-BRIEFING.md to match.
Split horizon C2, attack infrastructure: two front doors that do not share a fate, Apache fronting Sliver and Nginx fronting Mythic and Adaptix, each redirector on its own peered network, with the teamservers, collector, and operators behind a jumpbox.

The two errors in the top bar are the canvas validating as you build: the redirectors have no domain yet, which you set to one you control before deploying. It is caught here, not at apply time.
Multiple operators, one stack. An attack-infrastructure jumpbox can name a
list of operators (handle plus role), and each gets a Guacamole portal login
on the shared lab password. Set the jumpbox's access_mode to wireguard or
openvpn (default is a public portal) and every operator also gets a personal
VPN credential, generated on the jumpbox during provisioning: the keys never
enter the export, and each operator's client config, which holds their private
key, waits on the jumpbox for you to fetch and hand over. The portal logins of
the operators named in the topology share the one lab password, so per-operator
isolation comes from each operator's own VPN credential, not the portal login;
individual portal passwords for them are on the roadmap. On a VPN access mode the
portal closes to the internet and moves behind the tunnel, while SSH stays open
so the admin can keep deploying and managing the box, and
sudo rsp-operator add <handle> (also remove and list) adds a teammate to
the running jumpbox, printing their VPN credential and a portal password of
their own. See the
wiki Deploying a Range.
Harbor, a target range: a small corporate forest, a root domain and a child over a parent-child trust, with the ordinary path from a phished workstation to the forest.

GOAD, the full lab: three domains across two forests, five machines and their trusts, the reference range the written solution follows.

[!IMPORTANT] The export is the boundary. The canvas generates files; you run them under your own credentials. redStackPRO never deploys anything and never holds your cloud credentials or keys.
[!CAUTION] Authorized use only. redStackPRO builds offensive infrastructure and deliberately vulnerable ranges. Use it only in lab environments you own or are explicitly authorized to test, never against systems you do not have written permission for.
Pre-release, with more to come before 1.0. The pipeline itself works end to end: a topology compiles to Terraform and Ansible, and the export deploys.
| Provider | State |
|---|---|
| GCP, AWS | Supported and tested end to end, for both target ranges and attack infrastructure. |
| Azure | Supported and tested end to end, for both target ranges and attack infrastructure (new in 0.10.0). |
| Proxmox, ESXi | On the roadmap, not yet supported. |
The whole canvas in one container, the API and the web app on one port:
docker compose up # http://127.0.0.1:8000
docker compose up --build # build the image from this checkout instead
Compose runs the image it pins, ghcr.io/devzero-security/redstackpro:0.10.0,
pulling it from GHCR and building it from this checkout only when the pull fails.
Or run the published image without compose (:latest and :0.10.0 are the same
build):
docker run -p 8000:8000 -v redstackpro-data:/data \
ghcr.io/devzero-security/redstackpro:latest
The canvas listens on 8000 inside the container. To serve it on a different host
port, change the left half of the mapping (-p 8787:8000), or set
REDSTACKPRO_PORT for compose (REDSTACKPRO_PORT=8787 docker compose up in bash,
$env:REDSTACKPRO_PORT=8787; docker compose up in PowerShell).
Compose also carries an optional Postgres backend for a shared deployment:
REDSTACKPRO_DATABASE_URL=postgresql+psycopg://redstackpro:redstackpro@db:5432/redstackpro \
docker compose --profile postgres up
The image is the composition layer only. It does not carry Terraform or Ansible and never holds your cloud credentials: you run the export it produces from your own machine, exactly as in the from-source flow below.
Python 3.11 or newer, and Node 24 for the canvas.
git clone https://github.com/devZero-Security/redStackPRO.git
cd redStackPRO
python -m venv .venv
. .venv/bin/activate # bash (Linux, macOS)
.venv\Scripts\Activate.ps1 # PowerShell (Windows)
pip install -e ".[dev]"
The canvas is two processes, the API and the web app. Start the API:
redstackpro serve # the API, http://127.0.0.1:8000