Your first run in 10 minutes.

1. Unpack and create .env

Unpack the ZIP, open a terminal in the unpacked folder and create the configuration from the template:

cp .env.example .env            # Windows: Copy-Item .env.example .env

Set your own admin password in .env:

SOFTWAREFABRIK_ADMIN_PASSWORD=your-own-password

Nothing else needs changing for the mock walkthrough: you pick the mock in the first-time setup in a moment, and without a license a run started from the interface runs with the mock anyway — announced before you click. You need no agent CLI for it. The .env holds credentials: do not share or commit it.

2. Start and sign in

./start.sh                      # Windows: .\start.ps1

The script checks the requirements, starts the database in a container and then the factory. Open http://localhost:8080 and sign in as admin with your password. Later you stop the factory with Ctrl+C in the terminal; docker compose down stops the database, the data is kept.

3. Initial setup

As administrator you see “Willkommen — richte die Fabrik in drei Schritten ein” (welcome — set up the factory in three steps) on the dashboard. Click “Ersteinrichtung starten” (start initial setup):

1. KI-Zugang (AI access): later you choose Claude Code & co. and subscription or API key here. For this walkthrough click “Erst ohne KI-Agent ausprobieren (Mock)” (try without an AI agent first) at the bottom: the mock becomes the default — a run shows the whole flow but produces no real code. You switch to a real agent later under Einrichtung → KI-Zugang.
2. System: the system check shows only open points, each with a next step; it lists the mock as the active adapter. For the mock none of them is an obstacle — not even the “Lizenz” (license) row: without a license no new run with a real AI agent starts; runs with the mock remain possible. You only need the free Community license (“Lizenz einrichten”, set up license) for real runs. Then click “Weiter” (next).
3. Loslegen (get started): “Neues Projekt beschreiben” (describe new project) opens the quickstart.

Top right, “Einfach | Experte” (simple | expert) switches how much the interface shows. A new factory starts in “Einfach”; this guide follows that view.

4. Describe and build

The quickstart needs three inputs:

1. Project name: My first backend
2. Kind of project: Web application with Java (Spring Boot)
3. What should be built: A Maven project with Spring Boot, a health endpoint and a first smoke test.

localhost:8080/wizard/schnellstart

Above the button it says that the run uses the mock and produces no real code (the mock costs nothing either). Click “Bauen” (build; without a license and without choosing the mock it reads “Mit dem Mock ausprobieren”): the factory creates the project and starts the first run right away — your description is its task. You land on the run page; the “Mock-Lauf” box reminds you that this is a trial run.

5. Run page: watch and approve

The run works in the background; the page updates itself. The box “Was jetzt?” (what now?) tells you what to do in every state, and only the buttons that work right now are shown.

localhost:8080/runs/…

1. Before execution the run stops in Wartet auf deine Freigabe (waiting for your approval). Look at the changes and click “Freigeben und fortsetzen” (approve and continue).
2. Because the mock writes no build files, the build check reports “Build nicht geprüft” (build not checked) — a warning, not an error. The run waits a second time, this time before the merge. Approve again; it then runs to Abgeschlossen (completed).

If something goes wrong, the cause is shown at the top; “Erneut versuchen” (try again) starts a new run with the same task.

6. Result

After completion the “Ergebnis” (result) card sums up where the changes are and which files changed; under “Selbst ausprobieren” (try it yourself) is the command to start the application. “Ergebnis als ZIP herunterladen” (download result as ZIP) gives you the project folder without Git data and build artifacts. Under “Logs anzeigen” (show logs) you see every line of the run. The mock only simulates the run and writes no code; it skips the AI reviewers in the quality gate (the run page and the log say so).

The next 20 minutes: explore

• Project page: description, “Nächster Auftrag” (next task) with “Bauen”, latest result and previous runs.
• Artifacts: switch to “Experte” and open “Generierte Artefakte” (generated artifacts): PROJECT.md, INSTRUCTIONS.md, AGENTS.md. You see how your description became a task for the agent.
• Keep improving: on the result card, “Weiter verbessern” (keep improving) leads to the next task; “Vorschläge machen lassen” (get suggestions) starts a plan run that puts suggestions into the backlog.
• Detailed wizard: if you want to choose stack and checking tools yourself, try the six-step wizard in “Experte”.