Getting Started — Software Factory Docs

System Requirements

Requirement	Minimum	Recommended
Deployment	Docker, or the `sfactory-code` single binary	Docker on a machine that stays on
Coding agent CLI	One of Claude Code, Codex, or Forage installed and reachable	Whichever CLI you already use day-to-day
RAM	512 MB for the orchestrator	2 GB+, plus whatever the coding-agent CLI needs

Installation

Software Factory is self-hosted — run it with Docker Compose, or download the single self-contained sfactory-code binary. No hosted service to sign up for.

Download the sample docker-compose.yml from the Download page:

docker-compose.yml

services:
  sfactory-code:
    image: squrl-labs/sfactory-code
    ports:
      - "4287:4287"
    volumes:
      - ./data:/data
    restart: unless-stopped

Then start the stack:

Terminal

docker compose up -d

Download sfactory-code for your platform from the Download page.
Make it executable (macOS/Linux) and run it:

Terminal

sfactory-code --start

Once running, the web UI is available at:

Web UI

http://localhost:4287

Connecting a coding-agent CLI

Software Factory is an orchestrator around external AI coding-agent CLIs — it does not write code itself. From Settings → Coding Agent, choose which CLI the implement stage should drive:

Claude Code
Codex
Forage — Squrl Labs' own knowledge-graph-backed coding assistant

Whichever CLI you choose must be installed and reachable from wherever Software Factory is running.

Connecting a repository

Open the web UI, go to Repositories, and connect a GitHub or Squrl Nest repository. Software Factory needs access to read and file issues, and to open pull requests against it.

Start a new repo in Supervised mode so you can see each pipeline stage before it runs unattended. Switch to Auto once you trust the pipeline's output on that repo.

Next steps

Read The Pipeline to understand what happens to an issue from filing to pull request, and Bug Sweeps to see how Software Factory finds bugs nobody has reported yet.