Open source · Self-hosted · Yours to audit

Run AI and automation on your own hardware.

We build and document the stack that keeps inference, workflows, and data on machines you control — 120GB of unified VRAM on a single AMD APU, n8n pipelines you can import in a minute, and developer tools that never upload what you paste.

17
browser-based dev tools
6
importable n8n templates
120GB
unified VRAM, documented
zypher@halo:~
$ rocm-smi --showmeminfo vram
VRAM Total: 122880 MB
VRAM Used : 71344 MB
$ systemctl status lemonade
active (running) — local LLM gateway
$ n8n start
Workflow engine ready on :5678
$ _

Built for people who want to own their stack

Solo-led since 2021, with collaborators. Focused on open source AI, automation, and tools that stay under your control.

Open source at the core

Zypher Systems designs workflows and software for local AI inference, n8n automation, and practical DevOps utilities. We prioritize transparency, self-hosting, and code you can audit.

Day-to-day that means Strix Halo and Lemonade Server on Fedora, ROCm-friendly setups, and workflow libraries you can adopt without a vendor lock-in tax.

More about how we work →

Automation first

Workflows that cut busywork

AI native

Local and cloud models

Open source

Transparent by default

Production ready

Built for real use cases

stack.yml
hardware: Ryzen AI Max+ 395
os: Fedora / Linux 6.16+
inference: Lemonade Server, llama.cpp
runtime: ROCm 6.x
automation: n8n, self-hosted
data: PostgreSQL 16
delivery: Docker Compose, GitLab CI
telemetry: Umami (self-hosted)