Sypha AI Docs
Syphax

Core Capabilities

AI Agent Swarm Architecture, Sandboxed Execution Environment, Target Support, and Deep Scan Assessment.

AI Agent Swarm Architecture

Syphax runs a hierarchical multi-agent system rather than a single AI model. When a scan starts:

  • A root agent coordinates the overall engagement
  • Specialist sub-agents are spawned for each application component (auth system, payment gateway, admin panel, API layer, etc.)
  • Each component agent further spawns vulnerability-specific agents (SQLi agent, XSS agent, SSRF agent, auth bypass agent, etc.)
  • All agents run concurrently, maximising coverage and minimising scan time

This architecture means Syphax scales horizontally — complex applications get more agents, not slower scans.

Sandboxed Execution Environment

Every scan runs inside an isolated Docker container (Kali Linux base) provisioned per scan. The sandbox includes:

  • Full offensive security toolkit: Nmap, Nuclei, httpx, Gobuster, Arjun, TruffleHog, Semgrep, Trivy, Wapiti, ZAP, Retire.js, and more
  • Caido web proxy for intercepting and replaying HTTP traffic
  • Playwright for browser-based testing of JavaScript-heavy applications
  • Python scripting environment for custom exploit development
  • Automatic cleanup after scan completion — no state leaks between scans

Target Support

Target TypeExamples
Web applicationhttps://app.example.com
APIhttps://api.example.com/v1
GitHub repositoryhttps://github.com/org/repo (cloned into sandbox)
Domainexample.com (with subdomain enumeration)
IP address192.168.1.1 (with port scanning)

For repository targets, Syphax clones the codebase into the sandbox workspace and performs whitebox analysis — tracing code paths, reviewing authentication logic, identifying secrets, and testing business logic from the source.

Deep Scan Assessment

Syphax runs an exhaustive, multi-phase engagement across five phases. See Deep Scan Assessment for full details.

On this page