We have written about what the factory does and why we designed it this way. This post is for the people who want to know how it works — the compilation pipeline, the prompt assembly, the engine configuration, and the security model that keeps everything contained.
We have seen what happens when teams bolt AI onto their workflows without guardrails. You get a firehose of speculative issues, vague suggestions, and noisy PR comments that train developers to ignore everything the bot says. The Elastic AI Software Factory is built on the opposite philosophy: an agent that has nothing useful to say should say nothing at all.
What does it look like when a repo runs the full Elastic AI Software Factory? We deployed the framework on our playground repository -- combining the named workflows that ship with ai-github-actions and a set of custom Scheduled Audit configurations (Ideas Man agents, Playwright smoke testers, and domain-specific analyzers). The result is over 30 agents running on a daily schedule, and from the outside it can look like chaos. But there is a rhythm to it.
Code changes. Docs don't. These agents make sure they stay in sync.
These agents guard the things developers forget to check.
Five silent sentinels run every weekday morning. You never hear from them -- unless something breaks.
What if your repo generated its own feature backlog every morning? Not vague "we should consider" suggestions from a brainstorming doc nobody reads, but concrete, codebase-grounded feature proposals filed as issues -- five every day, each from a different domain expert. That's what the Idea Machines do.
Most agents find things. The Bug Hunter is coached to find nothing -- and considers that a success. Its job is to hunt for reproducible, user-impacting bugs every single weekday. Most days it comes back empty-handed, and that's the best possible outcome. A healthy codebase doesn't need a bug report. The Quality Crew is built around that philosophy: find real problems, fix them concretely, and stay silent when there's nothing worth saying.
A new issue lands. Within seconds, it's labeled, prioritized, and checked against every existing issue for duplicates. No human triaged it. No one assigned a label from a dropdown. The Issue Squad handled it — seven agents that collectively own the entire lifecycle of a GitHub issue, from the moment it's filed to the moment it's closed.
CI is red. The logs are 4,000 lines of webpack noise. Somewhere in there is the one line that matters. That's the job of the Detectives -- three agents whose entire purpose is to read failure logs so you don't have to.