One subscription · both products

Cygnus, end to end.

A single license unlocks both the hosted 1.5B steering API and the local 32B desktop runtime. Use either, or both.

Pro Lite · hosted
Qwen-2.5-1.5B-Instruct steering over HTTPS. No install, runs on any 4 GB consumer GPU. Just call the endpoint.
Activate Pro Lite →
Pro 32B · local
Desktop AppImage with the full 25-probe Patent VII adapter on Qwen-2.5-32B. 24 GB GPU recommended.
Download AppImage →
Local 7.5.1 · Governance Edition
Full proprioceptive 32B build with Constitution v1.0, 6-phase reasoning pipeline, 100M-token BM25 RAG, prompt-injection firewall, encrypted audit log. Download .tar.gz →
$39 / month · 7-day free trial · cancel anytime
Start 7-day trial →

Already a customer? Sign in to activate →  ·  downloads below require no signup.

Card required to start the trial. Your license key arrives by email within seconds of checkout and works for both products immediately.

CYGNUS Pro · v1.0.0-rc.7

Download CYGNUS

Native installers for macOS, Windows, and Linux. Five minutes from click to first response.

After install

Three steps to first response

i.

Launch & wait for the model

First boot loads Qwen-2.5-3B (~ 60 s on warm cache, ~ 90 s on cold). The status badge shows live progress. The 11 v6 probes finish loading next.

ii.

Activate (or stay in trial)

Settings → License → paste your activation key from the email we sent. Or skip — the trial gives you 7 days of full chat with the local probe-monitored model.

iii.

Send something to the model

Watch the 11 probes fire on the dashboard at ~ 10 Hz. Every response is QDKS-verdicted and recorded in the local audit log.

Requirements

System requirements

Mock mode (default)

  • 4 GB RAM
  • 2 GB free disk
  • No GPU required
  • Network only for license + auto-update checks
  • Internet for activation; offline thereafter for 7 days

Real-model mode (optional)

  • NVIDIA GPU with compute capability 7.0+ + ≥ 8 GB VRAM
  • CUDA 12.1+ runtime, recent driver
  • Python 3.10 with torch + transformers + bitsandbytes
  • ~ 6 GB free disk for the 4-bit Qwen-2.5-3B weights
  • v1.1 will install these for you in a managed venv