Goes live November 24, 2026

It appears on the home page and opens for votes when its launch week starts.

openfabriq

openfabriq

Your AI software factory, as code

FreeAICodeAI Agents10 impressionslaunches Nov 241 comment

Comments

>log in to comment
  • Matthieu Guirlinger[maker]· 2d ago

    Hi, I'm Matthieu, a freelance product engineer in France. I built openfabriq so a team can set up one shared factory on a repo, simply, and run it on its own infra. The agents, their triggers and what they may do live in the repo; the engine runs on your machine or your team's server, not on GitHub Actions runners. Two things I can show today: - This repo runs its own factory. The triager labels every new issue, and some of the merged PRs were opened by the factory. - On a test app, a labeled issue became a reviewed PR in 3.5 minutes, for $0.51 in API cost. It's open source (Apache-2.0) and built with Claude Code. I'm around all day to answer questions. What's the first task you'd hand to an agent on your repo?

openfabriq is a software factory as code: the AI agents that work on your repo (triage, implement, review) are defined in a folder of that repo, shared by the whole team. - On your infra: the engine runs on your machine or server, not on GitHub Actions runners. The model is Claude, through Claude Code or an API key. - Triggered by your tools: GitHub events, schedules, Slack, Linear, Sentry, or any signed webhook. - Versioned and planned: you change the factory config through pull requests, like your CI. "openfabriq plan" shows what a change does, new permissions first; "openfabriq apply" deploys it. - No keys for agents: an agent never holds a GitHub token. It proposes a PR, a label or a comment, and the engine applies only what the config allows, after running your tests. - Reviewed before GitHub sees it: no agent approves its own code. A separate reviewer, on another model if you want, reads each change before the PR exists. A person merges. - Visible: a studio shows each run like a Claude Code session, with its cost, a board and the workflow as a graph. - Documented: an optional wiki maps your code. The engine keeps it current through PRs, and agents read it for context. To try it: brew install openfabriq/tap/openfabriq, then openfabriq init in your repo; Claude Code writes the factory with you. Or start from one of 28 templates. v0.1, Claude only, macOS and Linux.

More AI tools →80 free dev tools →

  • AI agents defined as code in an openfabriq/ folder
  • Plan and apply workflow similar to Terraform
  • Runs locally or on your own server with Claude Code
  • Local Studio app to monitor factory runs
  • Built‑in templates for CI doctor, bug reproducer, test coverage

openfabriq is an AI-driven software factory that runs agents on your own machine to automate repo tasks via code-defined configurations.

for
Dev teams that want self-hosted AI automation for GitHub repos.
pricing
open source
license
Apache-2.0
openfabriq/openfabriq 1 0TypeScriptupdated 2 days ago
works withGitHubSlackLinearSentryany signed webhookHomebrewnpm

Key features

6 features of openfabriq
  • Factory as code — Define agents, triggers, and permissions in an openfabriq/ folder versioned with your repo.
  • Local execution — Runs on your macOS/Linux machine or server, not on external CI runners.
  • Secure permissions — Agents never hold GitHub tokens; the engine enforces config-based write limits.
  • Review workflow — Changes are reviewed by a separate reviewer agent before any PR is created.
  • Studio UI — Visual board shows runs, costs, and workflow graph for monitoring.
  • Template library — Start from 28 ready-made factories for common tasks like triage or CI doctor.

Use cases

  • Automatically triage new GitHub issues and post initial answers
  • Generate and merge dependency update PRs with safety checks
  • Fix production errors from Sentry by creating regression tests and patches
  • Run CI doctor to analyze failed checks and suggest fixes
  • Create advisory code reviews on every pull request

openfabriq vs alternatives

openfabriqn8nLangChainLlamaIndexDevin
Best forSelf-hosted AI agents for repo automationGeneral self-hosted workflow automationLLM app development frameworkData indexing for LLMsAutonomous AI coding assistant
PricingOpen sourceSubscriptionFreeFreeSubscription
DevHunt upvotes10000
LaunchedNov 2026————
  • openfabriq vs n8n: n8n is a general workflow automation platform, not focused on AI agents running on your own infra.
  • openfabriq vs LangChain: LangChain provides a framework for building LLM apps, but does not include the built-in factory-as-code and local execution model of openfabriq.
  • openfabriq vs LlamaIndex: LlamaIndex focuses on indexing data for LLMs, whereas openfabriq orchestrates AI agents to act on code repositories.
  • openfabriq vs Devin: Devin is an autonomous AI software engineer, but openfabriq offers team-wide configurable agents with versioned factory files.

openfabriq FAQ

Where does openfabriq run?+

It runs locally on macOS or Linux machines, or on a server you control.

Which LLM does it use?+

It uses Anthropic Claude (Claude Code or an API key) as the underlying model.

How are agents authorized to write to the repo?+

Agents have no GitHub token; the engine applies only the actions allowed in the factory config after running your tests.

Can I preview changes before they are applied?+

Yes, `openfabriq plan` shows the proposed changes and permissions; nothing is applied until you run `openfabriq apply`.

Do I need to install additional tools?+

On macOS you need Homebrew; on Linux you also need bubblewrap, socat, and ripgrep for sandboxing.

Is openfabriq open source?+

Yes, it is released under the Apache-2.0 license and can be installed via Homebrew or npm.

Summarized by DevHunt from openfabriq.dev · Nov 17, 2026. Details may change; check the official site.