Notes

• Describe your requirements in non-technical language - for example, "block American Express" - without using technical commands or editing configuration files.
• Tells you exactly what is left before you can go live. At the end, it lays out the remaining steps in plain terms, like switching from test to live account details, or pointing the app at your real backend instead of the example one.
• The builder saves information about your application, so it resumes your project exactly where you left off, in case a new session was started.
• When working in Claude Code, extra guardrails called hooks can run alongside the agent for safety - they stop Nuvei account details from ever being pasted into the chat, written to a file, or read back out, even by accident.
• The example backend the builder can generate uses Python. The same endpoint can be reimplemented in any language the production backend runs in - the app only validates that the endpoint exists and answers correctly.

warning

Not implemented yet.

• Work that produces large outputs is delegated to subagents, which keeps the main context window small and makes the agent less likely to hallucinate.