Autoattack

Deploy your first AutoAttack campaign.

Give the autonomous red team agent a goal and an authorized scope. Run its Docker command on a dedicated Linux host, then follow the attacks it chains at machine speed.

Contact Jump to the command

Before you start

Use a dedicated deployment host, separate from the systems you are authorized to test.

Linux and Docker
Full root access and rootful Docker. Rootless Docker and the container option --user are unsupported.
Network access
The host must reach your target systems and allow outbound HTTPS on port 443 to agent.autoattack.ai. The container uses the host’s network.
Authorized scope
Set the objective, authorized scope, exclusions and starting access before deployment. Review the reversible changes included in the campaign setup.
Starting access
Begin without credentials, or supply an account to test what an attacker could reach if that account were compromised.

Deployment host

Linux + rootful Docker
Campaign container

Your scope

Authorized systems
Named objective

The container has a read-only root filesystem and writable temporary storage. It performs the assessment against your authorized scope. Docker removes it when its process exits.

Review the campaign controls

Run your first campaign

  1. Contact

    Use your email address to sign in or create an account, then create your first campaign.

  2. Set the objective and scope

    Name the goal the agent should pursue, such as Domain Admin on your corporate domain. Review the authorized scope, exclusions, starting access and reversible changes.

  3. Deploy

    Run the provided command on the dedicated Linux host. The campaign begins when the agent connects.

  4. Read the route

    Follow the route, access kinds and weakness labels as the campaign progresses.

  5. Reveal the evidence

    Read the narrative with anonymous target labels for free. Subscribe to see target names, captured evidence and remediation advice. Checkout uses the discovered host count.

How do I choose an objective?

Name the goal the agent should pursue, such as “Read the CEO's inbox that's in our on-prem Exchange Server.”, “Gain Domain Admin access on our full enterprise forest. Start out with this contractor account sanjay_temp:V!ct0riou5” or “Access customer data, start from the server at 10.10.20.4 and use the local account z.smith:SummerT!m3”. Identify the mailbox, domain or data store and include it in the authorized scope.

Starting access sets where the test begins. Without credentials, the agent starts from the deployment host’s network position. A supplied account lets it test what that account could reach. The objective guide covers other target types. The internal network testing guide works through the deployment segment, scope and success evidence.

The deployment command

Use the command from your campaign page, where the token is already filled in. This example shows its structure and stops unless the shell variable AUTOATTACK_CAMPAIGN_TOKEN contains your campaign’s token.

Docker on Linux
docker run --pull always \
  -e AUTOATTACK_API_URL=https://agent.autoattack.ai \
  -e AUTOATTACK_CAMPAIGN_TOKEN="${AUTOATTACK_CAMPAIGN_TOKEN:?Use your campaign token-filled command}" \
  --stop-timeout 120 -d --rm --network host \
  --cap-drop ALL \
  --cap-add NET_RAW --cap-add NET_BIND_SERVICE \
  --cap-add NET_ADMIN --cap-add SYS_TIME \
  --read-only --tmpfs /tmp --tmpfs /run:exec \
  autoattack/agent:latest

Scroll the example sideways to read long lines.

Keep the campaign token private. When the agent connects, the waiting page opens its activity. You can leave that page and return through Campaigns.

Back to contents

Connection help

A waiting campaign has not received the agent’s connection. Check the deployment host before running another copy of the command.

Docker returns an error

Read the error in the terminal. Run docker info to confirm Docker is running and your account can access its rootful daemon. If pulling the image fails, check the host’s access to Docker Hub.

The command returns a container ID

Use docker ps to check whether that container is running. While it exists, docker logs CONTAINER_ID shows its startup output. Replace CONTAINER_ID with the ID returned by your command. Docker removes the container after it exits because the command uses --rm.

The container runs, but the page still waits

Run this check from the deployment host:

curl --head --connect-timeout 5 https://agent.autoattack.ai/healthz

An HTTP response confirms that the host can resolve the address and reach it over HTTPS. A resolution error points to DNS; a connection or certificate error points to the network or TLS configuration.

Confirm that the deployment command belongs to this campaign. Reissuing it revokes the previous token, so use the current command from the campaign page.

If the problem continues, email hello@autoattack.ai with the campaign identifier, visible status and relevant error. Remove the token and any credentials from messages and log excerpts.

Back to contents

Reading the report

Start with the access gains and their times, then follow the steps that produced them.

reached
The campaign established the access your objective required.
results
The summary shows access counts and gain times alongside the objective conditions still open.
not applicable to the systems assessed
The completed assessment found no systems matching the kind of target your objective required.

While the agent works, the campaign shows its running status. Stopped or failed describes an interrupted run. Check the campaign’s status and explanation before starting a new campaign.

Each step describes an access or credential gain and the steps that enabled it. Open Evidence and fix beneath a step to inspect its captured output and remediation advice with a subscription. Free campaigns keep the narrative and weakness labels visible with anonymous target names.

Use Print report on the campaign page to print or save it as a PDF. You can request a stop from that page. After a change to your network, create a new campaign with the objective and scope needed to test the affected systems.

Read the report guide
Back to contents

Create your first campaign.

You can review its scope and deployment command before starting the agent.

Contact
Autoattack