docker-compose-audit

docker-compose-audit

Audits docker-compose.yml files for security misconfigurations, providing severity ratings, remediation text, and YAML fix snippets for AI agents.

Category
访问服务器

README

Docker Compose Security Audit

MCP server that audits docker-compose.yml files for security misconfigurations. Trivy-grade check catalog, designed for AI agents — every finding ships with a severity rating, full remediation text, and a YAML fix snippet you can paste.

Built by Unbearable Labs. Pay-per-event pricing — you only pay when an audit runs.


Available on

  • Apify Actor Store — primary, metered usage (PPE)
  • MCPize — pending submission
  • MCP.so — pending submission
  • PulseMCP — pending submission
  • Smithery — pending submission
  • Glama — pending submission

Newsletter: Unbearable TechTips Weekly · All Actors: github.com/UnbearableDev

What it does

Point any MCP-capable client (Claude Desktop, Cursor, n8n, Make, Zapier, custom agents) at this server, hand it the contents of a docker-compose.yml, and get back a structured report with:

  • Severity — high / medium / low / info
  • Service — which compose service the finding affects
  • Description — what's wrong and why it matters
  • Remediation — what to do about it
  • Fix snippet — YAML you can paste directly into the file

Tools

Tool Purpose
audit_compose(compose_yaml? | compose_url?, min_severity='low') Run all checks, return full report
check_privilege(...) Container privilege & capability issues only
check_network(...) Network exposure issues only
check_filesystem(...) Volume mount & filesystem issues only
check_secrets(...) Secret hygiene issues only
check_resources(...) Resource limit issues only
check_image_hygiene(...) Image tag / registry / pinning issues only
check_runtime_lifecycle(...) Healthcheck / restart / init issues only
check_logging(...) Logging driver / rotation issues only
check_compose_hygiene(...) Deprecated fields / Compose-spec hygiene only
list_checks(category?) Browse the full check catalog

All audit-running tools accept the same input:

  • compose_yaml (string) — paste the YAML content directly, OR
  • compose_url (string) — public HTTPS URL to fetch (e.g. GitHub raw URL)

Provide exactly one. min_severity defaults to low (drops info findings); set to medium or high to filter further.

Example response (truncated)

{
  "summary": {
    "total_findings": 14,
    "by_severity": {"high": 3, "medium": 6, "low": 5, "info": 0},
    "by_category": {"privilege": 4, "network": 3, "secrets": 2, "...": 5}
  },
  "findings": [
    {
      "id": "DCS-002",
      "category": "privilege",
      "severity": "high",
      "service": "web",
      "title": "Privileged mode enabled",
      "description": "Service 'web' has `privileged: true`...",
      "remediation": "Remove `privileged: true`. If you need specific capabilities...",
      "fix_yaml_snippet": "    # remove `privileged: true`; if needed, use cap_add or devices selectively",
      "references": ["CIS-Docker-5.4", "NIST-800-190"]
    },
    ...
  ]
}

Pricing

Event USD
Any audit / check_* tool call $0.02
list_checks discovery call $0.005

You pay only when a tool is invoked. No subscription, no monthly minimums.

Check catalog (25 live in v1, growing toward 54)

Category Live checks
Privilege Root user (DCS-001), privileged mode (DCS-002), dangerous capabilities (DCS-003), cap_add: ALL (DCS-004), cap_drop: ALL missing (DCS-005), no-new-privileges missing (DCS-006)
Network network_mode: host (DCS-010), port bound to 0.0.0.0 (DCS-011), SSH port exposed (DCS-013), DB port exposed (DCS-014)
Filesystem /var/run/docker.sock mount (DCS-018), host root mount (DCS-019), sensitive host paths (DCS-020)
Secrets Hardcoded secret in env (DCS-026), secret-pattern env without Docker secrets (DCS-027)
Resources No memory limit (DCS-032), no CPU limit (DCS-033), no PID limit (DCS-034)
Image hygiene Unpinned / :latest image (DCS-037)
Runtime lifecycle No healthcheck (DCS-043), no restart policy (DCS-044)
Logging No log driver (DCS-048), no log rotation (DCS-049)
Compose hygiene Deprecated version: field (DCS-051), depends_on without healthcheck condition (DCS-052)

Use list_checks to get the canonical, up-to-date catalog with IDs, severities, and titles.

Connecting from Claude Desktop

Add to your MCP config:

{
  "mcpServers": {
    "compose-audit": {
      "transport": "streamable-http",
      "url": "https://YOUR-ACTOR-URL.apify.actor/mcp"
    }
  }
}

(Replace YOUR-ACTOR-URL with the Standby URL shown on the Apify Store page after you start the Actor.)

Limits

  • YAML size: 1 MB cap per audit call
  • URL fetch: 5-second timeout, max 3 redirects, HTTPS only
  • Session timeout: 5 minutes of inactivity

What's NOT covered (yet)

Pure static analysis of the compose file only. Out of scope for this version:

  • Image vulnerability scanning (use Trivy / Grype for that)
  • Live container inspection
  • Kubernetes / Helm manifests (different surface)
  • Dockerfile-specific lint (use Hadolint)

The next 29 checks on the v1.x → v2 roadmap include build-context security, additional capability checks, secret-pattern detection in build args, and registry trust verification.

Source / contact

Issues, ideas, or false-positive reports: open an issue on the GitHub repo or email unbearabledev@gmail.com.

get the weekly newsletter(https://unbearabletechtips.beehiiv.com).

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选