Projects

Git & AI Commits

PandaOS integrates Git directly into the workspace and uses AI to generate commit messages from real diffs, keeping your history clean without the effort.

Why integrated Git matters

When an AI assistant makes changes to your code, version control becomes more important, not less. You need to see exactly what changed, review diffs before committing, and maintain a clean history so you can roll back if something goes wrong. PandaOS puts Git directly in the workspace so that reviewing and committing AI-generated changes is part of the natural workflow, not an afterthought you handle in a separate tool.

Reviewing changes

The Git panel shows all changed files with full diffs. After Panda makes edits, you can inspect every modification before deciding what to commit. This is your safety net: AI-generated code should always be reviewed, and having the diff viewer right next to the chat makes review frictionless.

You can also switch branches, pull upstream changes, and push commits, all standard Git operations available without leaving the workspace.

AI-generated commit messages

Writing good commit messages is one of those tasks everyone agrees is important and nobody wants to do. PandaOS solves this by reading the staged diff and generating a clear, conventional commit message that accurately describes what changed.

The generated message is grounded in the real diff, not a guess. It summarizes the actual modifications: "Add input validation to user registration form" instead of "wip" or "fix stuff." You can use it as-is or edit it before committing. Over time, this keeps your Git history readable and useful for future you and your teammates.

Publishing your work

Once committed, push your changes to the remote to share them with your team or trigger CI pipelines. The full Git workflow, from staging through pushing, lives inside the workspace alongside the chat, file explorer, and terminal.

For working with files and running commands alongside the Git panel, see Files, Terminal & Browser.

Worktrees for parallel branches

When you want two branches in flight at once, ask PandaOS to work on something in a worktree. It creates a git worktree in its own folder, registers it as a linked PandaOS project, and switches you to it, so a long refactor and a quick fix stop taking turns on one checkout.

A fresh worktree from plain git is usually not runnable, because everything gitignored is missing. PandaOS hydrates the new worktree by copying the gitignored files the project needs (.env, credentials config, .mcp.json and similar) from the parent checkout, and can run your setup command afterwards. It only ever copies files git already ignores, and it asks before the first copy. If the list of what would be copied changes later, it asks again rather than quietly applying the new list.