If your team uses AI agents on its repo (one to triage issues, one to implement, one to review), running a single agent isn't the hard part. The hard part is turning them into a factory the whole team shares: one everyone can read, change and trust, instead of prompts and scripts living in one person's setup.
We solved this once already, for CI and for infrastructure: get the config out of people's heads and into the repo, versioned and reviewed like code. openfabriq does the same for AI agents.
Here's how to set one up with openfabriq: in a folder of your repo, versioned in git, and readable by people who don't write code.
The factory is a folder of your repo
The whole factory is an openfabriq/ folder: short Markdown and YAML files that say which agent starts on which event, and what it may do. One automation is one file:
--- when: issue.labeled: { labels: [ready-for-agent] } agent: implementer can: [create_pull_request, add_comment] --- Implement this issue and open a pull request. Triggers and rights are words from a closed list, so openfabriq validate checks every file, and a non-developer can read what the factory does.
Changing it is a pull request
openfabriq plan # what this change will do, new rights first openfabriq apply # deploy it to the engine plan compares your folder with what runs on the machine, like terraform plan. Run it on a pull request that touches openfabriq/ and you review the effects of the change, not only its text. The engine runs the applied copy, never a branch.
The whole team uses it
- Devs change the factory through pull requests and merge what agents propose.
- Product, support or ops start agents from the tools they already use: a GitHub label, a Slack mention, a Linear issue, a Sentry error.
- Everyone follows the work in the studio, and reads the code's map in the wiki (both below).
Agents hand work to each other through GitHub, the way a team does: a label starts the implementer, its pull request starts a separate reviewer, and a person merges.
The studio: what the factory does, for everyone
openfabriq start also serves a local studio. Each run reads like a Claude Code session, step by step, with what started it, what the engine applied or refused, and what it cost. A board shows what is waiting, working or needs a person, the workflow shows how work moves from one automation to the next as a graph, and each agent's page shows its rights and tools. A chat answers questions about the factory. On a team server, people sign in with GitHub. The studio only reads: work starts from GitHub, Slack, Linear, Sentry or openfabriq run.
The wiki: a map of the code, for people and agents
Turn it on with wiki: true in factory.yaml, and an agent that comes with the engine writes a map of the repository in short Markdown pages: where things live, how the pieces fit together, what the words mean. The pages live in your repo, in wiki/ by default, so every update arrives as a pull request you review. People read and search them in the studio, and agents read the pages they need before they work. It updates when you ask, or after each merge with refresh: after_merge.
On your own infra
The engine runs on your machine or a team server, with Claude Code or an Anthropic API key. Agents propose outputs, and the engine applies only what the config allows.
Try it
brew install openfabriq/tap/openfabriq # or: npm i -g openfabriq cd your-repo openfabriq init init lets Claude Code read your repo and write the factory with you, or you can start from one of 28 templates. The openfabriq repository runs its own factory: its issues are triaged by it, and some of its merged pull requests were opened by it.
It's v0.1, Claude only, macOS and Linux, open source under Apache-2.0. Feedback and contributions are welcome, and there are good first issues open.
openfabriq / openfabriq
Your AI software factory, as code: agents that triage, implement, review and fix your GitHub repository, described in readable files and run with Claude Code on your machine or your team's server.
Your AI software factory, as code.
AI agents that triage, implement, review and fix your GitHub repository,
described in readable files and run on your machine or your team's server
openfabriq in 85 seconds. Voice by elevenlabs.io · Music: The Warm Minimal by Pumpupthemind (Pixabay)
Quickstart · Documentation · Templates · Contributing
openfabriq runs a team of Claude agents on your GitHub repository: they triage issues, implement them and review pull requests. You describe them in an openfabriq/ folder of the repository, and the engine runs them on your machine or your team's server. Agents never hold GitHub credentials, and a person merges.
Quickstart
You need macOS or Linux, with Claude Code and the GitHub CLI logged in.
Install with Homebrew:
brew install openfabriq/tap/openfabriq Or with npm (Node.js 24+):
npm install -g openfabriq Then, in your repository:
openfabriq init # Claude Code reads the repository and writes openfabriq/ with… Docs: https://openfabriq.dev





