XCloak / docs

Recon Overview

XCloak Recon is an AI-orchestrated penetration testing platform — describe a goal in plain English, and it plans, runs, and reports a real security assessment using industry-standard tools. It's a separate product from XCloak Security Suite, currently in beta.

Recon is in active testing. Behavior, endpoints, and pricing described here may change before general availability.

What it does

Traditional scanners run one tool at a time and dump raw output for a human to interpret. Recon's AI planner reads your stated goal, decides which tools to run and in what order, executes each one in an isolated Docker container, then re-analyzes the combined output to strip false positives before scoring and reporting on what's actually there.

Every scan is a human-in-the-loop process — the AI proposes follow-up actions (e.g. deeper enumeration on a newly discovered port), but nothing beyond the initial plan runs without your approval.

The pipeline

Every scan follows the same eight-step, auditable sequence:

  1. You describe a goal — e.g. "Scan example.com for vulnerabilities"
  2. The AI Planner (GPT-4o, Claude, or local Ollama) breaks it into concrete tasks
  3. The plan is validated before anything executes
  4. Docker tool execution — nmap → nuclei → gobuster, etc., each in its own container
  5. AI analysis — validates findings, removes false positives
  6. CVSS risk scoring, with automatic stop conditions
  7. AI task proposals for follow-up work — held for your approval
  8. AI report generation, exported as PDF

Every step streams live over WebSocket, so you watch the scan happen rather than waiting for a final result.

Architecture

LayerRole
Next.js FrontendDashboard, new-scan flow, findings, schedules, settings
FastAPI BackendREST + WebSocket API, JWT auth, rate limiting
Execution EnginePlanner → validator → controller → parser → analyzer → risk engine → task proposer → report generator
Docker Workers7 isolated security-tool containers
Data & QueuePostgreSQL, Redis, and the configured LLM provider (Ollama / OpenAI / Claude)

Tech stack

LayerTechnology
FrontendNext.js 15, TypeScript, Tailwind CSS
DatabaseSupabase (PostgreSQL via Prisma)
StorageSupabase Storage
Scan EngineFastAPI, Python 3.12
Scan ToolsDocker — nmap, nuclei, gobuster, nikto, ffuf, sqlmap, whatweb
AIOllama (local) / OpenAI GPT-4o / Anthropic Claude — switchable at runtime
Threat IntelNVD API, AlienVault OTX
PaymentsRazorpay
Recon only scans targets you own or have explicit written permission to test. Target allowlisting is enforced at the platform level.