Skip to main content
Early Access. Commands and behavior can change between releases.
1

Open your project

Run architech in your project directory, then type:
Onboarding sets up your project locally; it deploys nothing to Cloud. The orchestrator, agents and skills ship inside Architech, so onboarding doesn’t copy them into your project.
2

Choose your project and board

Architech asks whether this is a new project or existing code, which GitHub Project board to use (or whether to skip it), and whether it may scan your repository to draft context and rules. It never scans without your permission. Review the proposed context before you accept it.
3

Review the proposed changes

Onboarding writes project memory, engineering rules, onboarding state and ignore entries. It never overwrites context you wrote.
Onboarding writes .architech/factory/factory.yaml (with mode: local, your board, repositories, label roles, issue naming and merge policy) and one pointer in CLAUDE.md.
Before writing the board setup, onboarding shows you a diff. Once you approve, it commits the factory file and the pointer together, and nothing else. If there is no CLAUDE.md, it creates one that holds only the pointer. It won’t write through a symlinked CLAUDE.md; follow the onboarding message to add the pointer yourself.You can run onboarding again later to add a board. Without a board, Architech still works for local software work.
4

Review the result

Read the onboarding report and do any manual steps it lists. If you picked a board, review factory.yaml: its repository list is the scope of board work, not every repository your organization owns.

Orchestrator and agents

Who plans, builds and reviews the work.

Factory quickstart

Start from your local factory file, then decide whether to move to Cloud.