OFFLIQ is an autonomous, 100% air-gapped AI engineer. Stop pasting proprietary client network data into public cloud APIs. Execute Windows server scripts locally with absolute zero telemetry.
OFFLIQ dynamically juggles models. A lightweight 'Gatekeeper' stays awake for instant chat. When coding, it loads the Heavy model, executes, and instantly terminates to free up 100% of system RAM.
Drop custom Python scripts (like Azure scanners) into the plugins folder. OFFLIQ locks them with a SHA-256 signature. If altered by malware, execution is permanently blocked.
Dynamic OS querying prevents junior techs from crashing their workstations. If an attempt is made to allocate more memory than physically available, the system intercepts and blocks it.
Drag and drop entire company IT manuals or codebases into the sources folder. OFFLIQ silently indexes them locally to write solutions customized to your specific internal standards.
Physically disconnect your router. OFFLIQ continues to code and execute perfectly offline.
A one-time CapEx license. Zero recurring SaaS fees regardless of how much code you generate.
Local Vector DB ingestion. Drop your historical SOPs into the folder for instant organizational memory.
Dynamic telemetry reads your physical VRAM and adjusts chunking to prevent OOM system crashes.
Provide a prompt or error log. OFFLIQ's Gatekeeper instantly queries your local Vector database for relevant company SOPs.
The Heavy Coder engine spins up, allocating dynamic VRAM to write the necessary PowerShell or Python script from scratch.
The Windows UAC shield triggers. Once a human clicks 'Approve', OFFLIQ executes the script natively on the host OS.
If the terminal returns a syntax error, the AI reads it, patches the local file, and re-runs the loop automatically.
Cryptographic Lock Active: The compiled binaries are secured. You must request a pilot key or procure a permanent license below to initialize the engine post-deployment.
V1.2 • ~72 GB
The complete air-gapped package. Includes the OFFLIQ Engine and all secure local Neural Matrices.
Download Master Setup~2.3 GB
Already have the models? Just download the updated execution engine.
Download UpdateDue to the massive 72GB encrypted size of the local neural matrices, standard browser scanners will flag the file as too large to scan. Please click "Download Anyway" on the secure routing page to begin the master transfer.
Request a free pilot, purchase a license, or contact the architecture team directly.
V1.2 Routing Agent