100% Offline · Zero Cloud · Zero Subscription

V.O.R.T.E.X

Voice Operated Recon & Tactical Exploitation eXecutor

You speak. It thinks. It scans. It reports.

A local-first, voice-and-text controlled penetration testing assistant — every command reasoned about, every scan executed, every result stored, entirely on your own machine.

session_id: vx-live-demo
4
Local Engines
0
Cloud API Calls
100%
Data Stays Local
Anatomy

Four senses. One local mind.

VORTEX is built around four modules, each standing in for a human sense — every one of them running entirely on your own hardware.

Ear
Whisper · Local
Converts spoken commands into text on-device. No audio ever leaves your machine.
Brain
Hybrid Rules + LLM · Ollama
Parses intent into a target and arguments — deterministic rules first, a local LLM as fallback.
Hands
Nmap · Subprocess
Executes the real scan against the real target — safely, with a stop switch always within reach.
Mouth
Live UI · SQLite · Export
Delivers results back — on screen, logged to a local database, exportable as a report.
The Science

How a command becomes a scan

Every instruction passes through a reasoning pipeline before a single packet is sent.

01 · Input

Voice or Text

“Fast scan localhost” — spoken or typed, captured and normalized.

02 · Rule Engine

Deterministic Match

Known patterns — port scans, OS detection, ping sweeps — resolve instantly, with zero ambiguity.

03 · AI Fallback

Local LLM Reasoning

Anything novel is handed to a local model, which infers a structured target + argument set.

04 · Execution

Real Scan, Real Result

Nmap runs the resolved command; the result is spoken back, logged, and made available for export.

Why hybrid, not pure-AI

Small local models can be inconsistent with structured output. VORTEX resolves common commands deterministically — instantly, with zero ambiguity — and calls on the AI only when a request doesn't match a known pattern. The result: the speed of rules, with the flexibility of reasoning when it's actually needed.

Interfaces

However you communicate, VORTEX listens

Two input channels are live today. Four more are on the runway.

Available
Keyboard Input
Type any command directly into the console.
Available
Voice Input
Speak a command — transcribed locally, no cloud round-trip.
Coming Soon
📄
File Input
Feed target lists or config files directly into a session.
Coming Soon
📷
Image Input
Point VORTEX at a screenshot of a network diagram or dashboard.
Coming Soon
📡
PCAP Input
Hand it a packet capture and ask what it means.
More channels
on the roadmap
Where This Is Going

From a single tool to an agentic operator

The next architecture doesn't wire the AI directly to a tool — it interposes a planner and a tool manager, so the reasoning layer stays decoupled from execution.

1User — Voice / Text
2Intent Understanding
3Goal Interpreter
4Task Planner
5Tool Selection Engine
Nmap
HTTPX
WhatWeb
Gobuster
FFUF
Nikto
Each tool runs behind its own adapter — the AI never touches a tool directly
6Result Correlation Engine
7AI Reasoning Layer
8Human-Readable Report
The enterprise leap

User → Analyze target → AI thinks → Chooses tools → Correlates results → Explains findings → Suggests next action. Every tool gets its own adapter (/tools/nmap.py, /tools/nikto.py…) behind a standard interface — so the AI is never wired directly to a tool. It reasons; the Tool Manager executes.

Technical Roadmap

Five phases, from proof of concept to platform

VORTEX is being built in deliberate stages — each one shipped, tested, and proven before the next begins.

Complete

Proof of Concept

The foundation: a fully working, fully offline voice-and-text pentesting assistant.

Voice input
Text input
Whisper (Ear)
Phi3 by Microsoft (Brain)
Ollama runtime
Nmap execution
Planned

Structural Foundation

Preparing the codebase to scale to many tools without rewriting the core.

Tool abstraction layer
Per-tool adapters
Standardized JSON output
Better logging
Session storage
Planned

Expanded Reach

More tools, chained workflows, and AI that summarizes instead of just reporting.

Multiple recon tools
Workflow chaining
AI-generated summaries
Planned

Autonomous Reasoning

VORTEX starts planning ahead, remembering context, and weighing risk.

Autonomous planning
Context memory
Risk scoring
Recommendation engine
Planned

Platform

From a personal tool to a shared, extensible security platform.

Plugin architecture
Custom modules
Report generation
Team collaboration
Ready When You Are

Your machine. Your scans. Your data.

No account. No API key. No usage meter. VORTEX runs where you run it — and nowhere else.