# Reefline

> Container image hygiene and runtime security for modern Kubernetes.

GitHub: https://github.com/siddhantprateek/reefline

Reefline gives security teams a simple, powerful way to scan container images for vulnerabilities, benchmark compliance, and layer inefficiencies — with AI-generated reports and an MCP server for agent-native access.

## What it does

- **Vulnerability Scanning** — Grype (Anchore). Detects CVEs across OS packages and language ecosystems with severity classification (Critical, High, Medium, Low).
- **CIS Docker Benchmark** — Dockle. Checks images against CIS Docker Benchmark best practices.
- **Layer Efficiency Analysis** — Dive. Breaks down each image layer, identifies wasted space and inefficient build patterns.
- **Image Inspection** — Skopeo-like metadata inspection via containers/image. No Docker daemon required.
- **AI-Powered Reports** — Supervisor + Critique multi-agent pipeline generates actionable security reports with Dockerfile optimization suggestions.
- **MCP Server** — stdio and HTTP modes. Compatible with Claude Code, Cursor, Claude Desktop, Archestra Chat.

## Architecture

Five components: UI (React dashboard) → API Server (Go/Fiber) → Worker (async scan jobs via Redis/Asynq) → Flow (Python AI report generation via LangGraph) → MCP Server (agent-native access).

Artifacts stored in MinIO: grype.json, dockle.json, dive.json, report.md per job.

## Integrations

- Container Registries: Kubernetes (in-cluster), Docker Hub, Harbor, GHCR
- LLM Providers: OpenAI, Anthropic, Google Gemini, OpenRouter
- LLM Proxy & Observability: Archestra AI (Tool Policies, cost monitoring, traces)

## Tech Stack

- Backend: Go 1.22+, Fiber v2, PostgreSQL + GORM, Redis + Asynq, MinIO, OpenTelemetry
- AI: Python, LangGraph, multi-agent (Supervisor + Critique)
- Frontend: React + TypeScript + Vite + shadcn/ui + Tailwind CSS
- Security: AES-256-GCM encrypted credential storage

## Author

Siddhant Prateek Mahanayak — Platform Engineer & Founder
GitHub: https://github.com/siddhantprateek
LinkedIn: https://linkedin.com/in/siddhantprateek