Skip to content

Quick Start

Get a server running, create an engagement, and choose your first test. The individual workflow runs without pairing or AI credits. Install Raven has every platform and option.

1. Get a server running

One-line install (easiest; Kali / Debian 12+ / Ubuntu 22.04+): 

curl -fsSL https://chamberdoorsecurity.com/install.sh | sh
raven-server start -p 8080 --no-browser
That adds the signed apt repository and installs raven-server, so Raven upgrades with the rest of your system (apt upgrade). Add the operator client with sudo apt install raven-cli.

Don't apt install raven

Debian ships an unrelated package called raven (Kali inherits it) whose version outranks ours, so that command installs the wrong tool. Ours are raven-server and raven-cli.

Docker (most platforms): 

gunzip -c raven-pro-*.tar.gz | docker load
mkdir -p ~/raven/pentests ~/raven/auth ~/raven/config
docker run -d --name raven-pro --restart unless-stopped \
  -p 127.0.0.1:8080:8080 \
  -v ~/raven/pentests:/home/raven/pentests \
  -v ~/raven/auth:/home/raven/.raven \
  -v ~/raven/config:/home/raven/.config/raven \
  --cap-add=NET_RAW --cap-add=NET_ADMIN \
  <image-tag>

No license key

Individual Raven needs no LICENSE_KEY and no paired account for its local core workflow. Pairing is optional for managed AI. Older images that require LICENSE_KEY should be replaced with the current release.

The Docker image bundles the shipped modules' tools, wordlists, templates, and local agent. Select an engagement to start the agent; tests run when you choose them. Keep the three mounts when upgrading. The separate auth/ mount holds the persistent credential key and optional pairing token; see Credential Encryption for backup and recovery.

Bare binary (toolchain already installed): 

# Linux (amd64 or arm64)
chmod +x raven-server-linux-* && sudo mv raven-server-linux-* /usr/local/bin/raven-server

# macOS (Apple Silicon). Clear the Gatekeeper quarantine flag first, or macOS
# kills the unsigned binary on launch. `xattr -cr` is a no-op when the flag was
# never set (a curl download never sets it), so it is safe to run either way.
chmod +x raven-server-darwin-arm64
xattr -cr raven-server-darwin-arm64
sudo mv raven-server-darwin-arm64 /usr/local/bin/raven-server

raven-server start -p 8080 --no-browser

Use the bare binary where the tools you need (nmap, nuclei, and friends) are already installed natively; unavailable tools cannot run. On macOS and Windows the Docker image above is the supported default.

2. Open the dashboard

Browse to http://localhost:8080.

The dashboard has no authentication in single-user mode, so where it listens matters:

  • Bare binary: binds loopback by default. Nothing else on your network can reach it.
  • Docker on macOS/Windows: the documented -p 127.0.0.1:8080:8080 publishes to loopback only. A bare -p 8080:8080 would publish on every interface instead.
  • Docker on Linux: optional --network host can let Raven scan your local network directly. That also means the dashboard listens on your network interfaces, not just localhost. On an untrusted network (client site, shared wifi) swap it for -p 127.0.0.1:8080:8080, accepting reduced host-network scanning.

For a remote or VM host, keep it on loopback and tunnel in: 

ssh -L 8080:127.0.0.1:8080 user@host

3. Create an engagement and choose a test

In the web UI, choose New Engagement, name your client, and add authorized targets to Scope. Open Testing, select the relevant assets and a check, review its targets and execution location, then run it. Supported results parse into hosts and findings. In Docker, the bundled agent supplies local execution without a jumpbox or in-app tool setup step. Target credentials, third-party service keys, and a network route still depend on the check you choose.

Use the separate Targets and Exclusions fields for scope; see the CIDR example. If the check must originate on another operator host, choose its execution path explicitly. The Remote Agent guide shows the binary, callback, server token, and engagement settings together.

For a worked example, follow Internal Lab: From Scope to Evidence. It covers discovery, SMB enumeration, isolated authentication capture, a manual finding, and a reviewed retest before reporting.

Review findings and evidence, then generate the basic report or export your work. This core workflow needs no AI.

4. Optional: use AI

Manual prompts: in AI Analysis, generate an anonymized prompt, review it, and copy it to the model you choose. Paste the response back for local placeholder restoration. No Chamber Door pairing or credits are required.

Native BYO analysis: open Settings → AI & privacy, choose OpenAI, Claude, Gemini, or a compatible API server, and save its model and credentials. Compatible API endpoints can include a local model. Prepare the engagement analysis in AI Analysis, review the exact prompt, endpoint, and model, then send it explicitly. Your provider account handles the request and any provider charges.

Managed AI: pair your Chamber Door account if you want the managed advisor, chat, finding drafts, or report writing. Click Pair this device, or run:

raven-server auth login     # bare binary / apt
# Docker: docker exec -it raven-pro raven-server auth login

Confirm at the printed URL. The token is saved under ~/.raven/ or the mounted Docker auth/ directory. Managed requests need available Chamber Door credits.

Raven replaces known identifiers and redacts recognized credential forms before AI requests. The mapping stays on the Raven host, and recognized response placeholders are restored there. Redacted secrets are not restored. Review the request: unfamiliar names and identifying technical context can remain. See AI Analysis for privacy limits and the differences between these three paths.

Next: Testing Workspace for target selection, workflows, execution, and retesting; CLI / REPL for the terminal; or AI Analysis for provider setup and privacy.