Working on your project

Working with Claude Code

Anthropic's AI agent in the terminal or as a desktop app – preconfigured for your platform projects.

On this page

Claude Code is an AI agent by Anthropic that works on your code directly in the terminal (or as a desktop app): it reads your project, writes changes, and runs commands.

Installation

The setup app installs Claude Code during developer setup (CLI and desktop variant). Afterwards you open projects directly from the setup app in Claude Code – or start the CLI yourself in the project folder:

cd ~/Documents/application-platform/<organization>/<project>
claude

Preconfigured for the platform

The project folder contains (symlinked from local-configuration):

  • CLAUDE.md – read automatically by Claude Code, pointing to the central rules.
  • AGENTS.md – the shared instructions for all AI tools: start applications with ap run-local, don’t change .gitlab-ci.yml or env/*.generated.env, put your own values into env/*.custom.env, never put secrets into commits or answers.
  • .claude/ – Claude-specific configuration.

Claude Code therefore follows the platform conventions without any extra setup.

Tip
Start Claude Code in the parent project folder, not inside a single repository – the agent then knows backend, app, and configuration at the same time.

Your own rules

How to give Claude (and all other AI tools) your own working instructions – e.g. committing and pushing after every task – is covered in Telling the AI how to work.