Compose ScanFind what can kill you before it reaches production.
Paste a docker-compose.yml and get a service-by-service security audit: privileged escalation, Docker socket mounts, host networking, hardcoded secrets, missing healthchecks, and a 0–100 score with inline copy-paste fixes for every finding.
- Security
- Docker
- DevOps
- Developer tool
Why it exists
docker-compose.yml files accumulate security debt silently. A developer adds privileged: true to fix a permission issue in dev, and it ships to production. Another mounts /var/run/docker.sock for a CI helper and never removes it. A third hardcodes a database password because the secrets manager was too slow to set up. These are not theoretical risks; they are the actual path attackers use to escape containers.
How it works
Paste the compose file
Drop the full contents of your docker-compose.yml into the input. The parser extracts every service block and runs the audit locally; nothing is sent to a server.
Review per-service findings
Each service gets its own risk card: name, image, risk level (CRITICAL / HIGH / MEDIUM), a 0–100 score, and every finding with a severity label, description, and the exact YAML fix to apply.
Apply the fixes
Copy the fix snippet directly from the finding; it's already formatted as the correct YAML. Add your AI key to generate a prioritized, step-by-step remediation plan across all services.
What it does
Critical escape-path detection
privileged: true, /var/run/docker.sock mounts, and cap_add: ALL are all CRITICAL, scored −35 each. These are the three most common container escape paths in the wild.
Hardcoded secret detection
Scans environment blocks for keys matching PASSWORD, SECRET, TOKEN, API_KEY, JWT, and similar patterns where the value is a literal string, not a variable reference.
Host network and root user flags
network_mode: host removes all network isolation. user: root means a compromised container has an easier path to the host. Both flagged HIGH with exact remediations.
Dangerous capability checks
SYS_ADMIN, SYS_PTRACE, SYS_MODULE, NET_ADMIN, and DAC_OVERRIDE in cap_add are individually flagged HIGH. SYS_ADMIN alone enables most published container escapes.
Hardening gap detection
Missing read_only: true, no-new-privileges not set, no healthcheck, and no resource limits are flagged as LOW or INFO: cheap to fix, high-value for defense-in-depth.
BYOK AI remediation plan
With your own key (Anthropic, OpenAI, Google, or Groq), generate a prioritized step-by-step plan: most critical fixes first, with exact YAML snippets. The AI receives finding titles and scores only.
Who it’s for
For backend developers, DevOps engineers, and platform teams: anyone who writes or reviews docker-compose.yml files and wants to catch security misconfigurations before they reach a staging or production environment.
Open & honest
- MIT licensed: every rule is readable in the source
- No upload, no backend: the compose file is parsed entirely in your browser
- AI call sends finding titles and scores only, no config values, no secrets
- BYOK: your key goes directly to your chosen provider, never to Aonami
- Three built-in samples to try without your own file: 100/100, ~80, 31/100
Scan your compose file.
Full service-by-service audit with copy-paste fixes runs with no key. Bring your own key for the AI remediation plan.