T3 Code review: macOS setup and CLI workflow comparison

Nuno installs T3 Code on macOS and reviews its desktop interface for an AI coding assistant workflow. He opens an existing Laravel project from a local folder; the project picker also offers a GitHub repository URL. Settings expose provider choices, appearance and key bindings, while session controls include model options, full access and Git worktrees.

The demo tests parallel sessions with separate questions about project dependencies. Nuno then requests a browser testing package and inspects command details, code snippets and session status. He compares the shared interface with his usual Claude Code CLI and OpenCode workflow, where he switches tools to use different models.

A custom action runs php artisan test in an embedded terminal. He also tries terminal splits and a side panel for files or uncommitted diffs. Commit message generation initially fails. Changing the text generation model to his Claude provider resolves the failure in this demo, and the button then creates the commit directly with an automatically generated title.

The usage panel shows estimates at API rates, which Nuno says do not represent his subscription charges. His positive assessment covers short demo sessions; he leaves performance with larger projects and longer parallel sessions untested. The walkthrough shows a desktop frontend, without establishing offline inference or the ability to run models locally.