Provision a dedicated VM (or equivalent host) with the specs below before running the worker. The worker container spawns short-lived sandbox containers per scan, so the host needs headroom beyond what the worker itself reserves.

​
VM requirements

ResourceMinimumNotes
vCPU8Worker reserves 2; the rest is for the host OS, Docker daemon, and per-scan sandbox containers running in parallel.
RAM16 GBWorker is capped at 4 GB; sandboxes can each use 1–2 GB during deep scans.
Disk40 GBOS + Docker images + scan artifacts. Artifacts persist in the tigerstrike_runs volume until the platform’s retention sweeper prunes them.
NetworkOutbound HTTPS (443)To your region’s API host (api.tigergate.dev on us1) and Docker Hub. No inbound ports required.

​
Operating system

  • Linux — Ubuntu 22.04 LTS, Debian 12, RHEL 9, or Amazon Linux 2023 (any modern x86_64 distro with kernel ≥ 5.10).
  • Root or sudo access for the initial Docker install.
ARM64 hosts work but are not formally tested.

​
Docker

  • Docker Engine 24+ (install guide)
  • Docker Compose v2 plugin — invoked as docker compose ... (not docker-compose)
  • The user running the worker must be in the docker group (or use sudo)
Verify your host before continuing: 
docker --version            # Docker version 24.x or newer
docker compose version      # Docker Compose version v2.x or newer
docker run --rm hello-world # confirms the daemon is reachable

​
Why these numbers

A pentest scan runs 2–4 parallel agents, each driving its own sandbox container with a browser, terminal, and HTTP proxy. On an undersized host the sandboxes start swapping, agent steps time out, and findings come back incomplete. The 8 vCPU / 16 GB / 40 GB shape gives clean headroom for two concurrent scans (the default --workers 2 setting in docker-compose.client.yml). If you plan to bump worker concurrency above 2, scale the VM linearly: each additional concurrent scan needs roughly 2 vCPU, 4 GB RAM, 5 GB disk.