olgac.org builds open-source tools for the people who keep servers running, and one for the living room. Each one starts from a real operational problem and does one job well.
$ systopsy analyze ./crash-2026-10-11 --ai claude › collected kern.log · journal · dmesg · app logs › redacted 3 secrets before analysis › timeline 14 events in 92 s before reboot ✗ crash OOM killer → mysqld (61.8 of 64 GiB) ! cause buffer pool sized above 85% of RAM ✓ fix cap buffer pool at 48G set vm.swappiness=1 evidence kern.log:18233, mysqld.err:902 ai: claude · also works with a local LLM, any AI API, or offline rules only
Three tools for server operators, plus one app for everyone. All open source, privacy-respecting and focused on doing one job well.
A small terminal app that checks the security of your Linux servers over SSH. It's a lighter alternative to Wazuh or OpenSCAP for teams with fewer than 50 servers.
A modern, open-source IPTV player for Windows. It's fast and clean, and it's careful with your provider's connection limits. You bring your own sources. LiberCast ships no content.
An autopsy for crashed servers. Systopsy collects logs, kernel messages and service errors from a failed Linux host, rebuilds the timeline and explains why it went down and what to change.
oneglassLightweight cyber asset attack surface management (CAASM). Oneglass compares your inventory, monitoring, security and virtualization tools and shows which servers are missing where.
Systopsy, our next release, turns raw crash evidence into a clear root-cause report. We build and test it with Claude by Anthropic first, and it also runs on a local LLM, any AI API, or no AI at all.
Gather kernel logs, journald, dmesg, service error logs and system metrics from the crashed host.
Local parsers build a timeline and cut gigabytes of logs down to the lines that matter, with secrets redacted.
The model you choose correlates events, ranks likely causes and cites the exact log lines.
You get a readable report: root cause, confidence level and concrete config or command fixes.
Our primary, best-supported provider, through the Claude API.
Bring your own key for another provider, including OpenAI-compatible endpoints.
Run a model on your own hardware, for example with Ollama or LM Studio. Nothing leaves your network.
Built-in rules catch common failures, and you can export a ready-made prompt to paste anywhere.
Credits let us make Claude the easiest way to use Systopsy and speed up how we build all four products.
One product at a time, built in the open. Two are already in public beta.
Watch the reposolgac.org is run by Veysel Olgac, a Linux system engineer who looks after production infrastructure every day, with a background in security, monitoring and alerting. He has mostly worked with open-source software, so building in the open comes naturally.
Every tool comes out of that work. Systopsy is for the crash whose answer is spread across five log files. nöbet is for small teams that need a security check but can't run a full security platform. Oneglass is for the VM that never made it into the inventory or the monitoring.
All code is public. Audit it, self-host it, contribute to it. Nothing is locked in.
Your data stays on your machines unless you choose to send it somewhere. No telemetry by default.
Designed around real production failure modes, not demo data.
AI makes the tools faster. Every core feature still works without it.
Everything is built in the open. Questions, ideas, or want to try an early version? Get in touch.
nöbet
Systopsy
Oneglass