Commander0.7.1046

Open a repository in Commander, select a branch or worktree, prompt an AI coding agent, inspect the resulting diff, and commit only the changes you approve. It brings Claude Code, Codex, and other supported agents into one Git workflow, letting you switch providers without editing configuration files or creating new accounts. Each agent run remains a reviewable unit, so you can understand a patch before it reaches the project rather than treating generated code as an unexplained commit.

Separate worktrees support three parallel tasks on isolated branches, making it practical to compare different implementations without manually juggling files. Toolbar actions can run project scripts such as builds, linters, and deployments, while chats can stay in tabs or move into separate windows. Sessions survive relaunches, and work or personal accounts plus local or cloud models can be configured once, assigned to projects, and switched instantly. Commander can also run agents on a remote Linux machine over SSH while keeping code review on the Mac, with VoiceOver support, a clear hierarchy, and keyboard navigation.

It suits developers who want an explicit review loop for risky refactors, parallel experiments, or any change where “the agent wrote it” is not enough, rather than an IDE or terminal replacement. The native Swift app is 30 MB and is not built with Electron. Commander is freeware in the Developer Tools category for MacOS, and version 0.7.1046 requires macOS 15 Sequoia or later on Intel or Apple Silicon.