Skip to main content
The Factory setup wizard is a form-based way to stand up a Software Factory. Instead of describing what you want to Smith in chat, you work through a sequence of screens, and the wizard creates the workspace, credentials, runtime environment, and agents the Factory needs. Each screen states one fact the next depends on, so the steps run in order and each passes its result forward. You can leave the wizard and come back: your progress is saved as a setup attempt, and the wizard resumes from the answers it already holds. For the conversational alternative, see Software Factory commissioning.

​
Steps

1

Account

Pick the organization or personal account that owns the workspace.
2

Workspace

Name the workspace your credentials and containers live in. One Factory, one workspace.
3

LLM Keys

The model credential the Factory calls, billed to your account.
4

Code

Connect the repository the Factory reads issues from and opens pull requests against.
5

Environment

The Factory scans your repository and writes a setup script to build it.
6

Agents

The four required agents that make up the pipeline.
7

Issues

Where issues live — usually already covered by the Code step.
8

Review

A last look at what is already true and what gets created when you confirm.
9

Live

Add the trigger label to an issue and the Factory gets to work.

​
Workspace

A workspace is where your credentials live and where the Factory’s containers run. Credentials attach to a workspace, not to an account, which is why the workspace comes before the model and repository steps.
Naming the workspace is a one-shot — a workspace cannot be renamed after it is created. Choose the name before you continue.

​
Code

Connect the repository through either the GitHub App or OAuth. The Factory reads issues from this repository and opens its pull requests against it.

​
Environment

The environment step turns the connected repository into a runtime environment the Factory’s agents can run in. The wizard scans the repository, proposes a setup script, saves it as a runtime environment, and then builds that environment to confirm it works. The scan reports its progress as it goes, so you can follow which part of the repository it is working through. You can read and edit the proposed script before it runs, and edit it again after a failed run. Run again tests the edited script.

​
When the test run fails

A failed run reports what it found: You can also open the session or the full container log from the same place. The build runs in the background, so a failure does not trap you in the step. You can continue setting up the Factory and return to the environment later — though the Factory cannot run until its environment builds.

​
Agents

A Factory runs on four agents: a planner, a coder, a reviewer, and a health agent. The wizard installs them for you.

​
Live

The Factory starts working when you add a trigger label to an issue. Labels drive the pipeline’s stages — planning, implementation, and review — and the Factory moves an issue through to a branch, a commit, a pull request, and a review.