Why FragonForge exists

In the winter of early 2026 I caught myself doing the same thing every evening: sitting in front of a laptop, watching a coding agent write better code than I do. Watching, because I could not trust it unattended. The quality was only as good as my attention that evening, nothing about it was proven, and stepping away felt like leaving a stranger alone with push access.

That obsession cost me things that matter more than code. My hobbies went first, then a relationship broke over it, and my cats saw more of my back than my face. Somewhere around February I stopped blaming my discipline and started blaming the shape of the tool: these agents are built to be watched.

So I built the opposite: a loop I can walk away from. Code quality that is checked statically instead of by vibes, with 33 analyzers, a gate that blocks new findings, and an A to F rating I can read without opening a single diff. Review as the only mandatory seat time. Nothing merges unless I explicitly allow it and the gate is green. If an agent writes better code than me, fine: then the least it can do is prove it while I am outside.

FragonForge was not meant to be a product. The first version was an internal tool for Fragon Studios, built for exactly one impatient user: me. It became a SaaS because nothing about the problem was unique to us. Every developer I know sits next to an open agent, watching.

We run it on ourselves. Our next product, FragonIntel, is being built with FragonForge today, and the next step is wiring the loop closed: errors from FragonIntel file labeled issues, and FragonForge attempts the fix. When we publish numbers from that, they will be real ones.

The mission has not changed since February: you should spend less of your life in front of a laptop. Including me. What that looks like in practice is on the use-cases page.

Danijel BalogFounder, Fragon Studios e.U.