Agents in parallel
Each issue gets its own agent and its own branch, so ten small changes move as fast as one.
Every change runs down the same line: intake, spec, build, verify, release. Agents work each station. Nothing ships until you merge.
Open source · Claude Code · Self-hosted
The line
Every change, from a typo to a feature, takes the same route. Each station has one input and one output, so you always know where a change is and what it is waiting for.
Why a factory
Each issue gets its own agent and its own branch, so ten small changes move as fast as one.
Acceptance criteria become tests before the code is called done. A criterion without a test isn't met.
Agents open pull requests; only you merge them. Production, secrets and releases stay in your hands.
The factory is built by the factory. Every issue, spec and pull request is on GitHub.
How it works
Point the factory at a GitHub repository and the place it deploys to.
Write one sentence. The factory turns it into an issue with acceptance criteria.
It comes back tested, reviewed and linked to its issue. Merge it and the release is cut.
Clone the repository and follow the README to start the factory on your own machine.
FAQ
A fixed line every change goes through: intake, spec, build, verify and release. Agents do the work at each station, and the line keeps the standard the same whoever, or whatever, does the work.
No. Agents open pull requests and fix what reviews and tests find. Merging, deploying and releasing are your call.
It is built around Claude Code today. Each station is a plain step with clear inputs and outputs, so other agents can take a station later.
Yes. It runs on your own machine or server with Postgres, and the code is on GitHub.
To your repository, on branches, through pull requests. Secrets stay in your environment and are never committed.