Checkpoints and undo
Commandr saves your project before agents change it, so any run can be put back — the whole run in one click, or any single checkpoint.
In short
- A checkpoint is saved before every pipeline run and every step that changes files.
- Undo the last run with ↺ in the pipeline bar, or from ⌘K.
- Before you restore, you see exactly which files change back.
- Your branch, staging area and git history are never touched.
When Commandr saves one
Commandr saves a checkpoint of the project folder before each pipeline run and before any agent step that changes files. Each one is named after what came next, so you can tell them apart.
To save one yourself, choose Save now on the pipeline board.
Undo the last run
Click ↺ in the pipeline bar, or press ⌘K and choose Undo last pipeline run. The whole run goes back in one step.
Restore any checkpoint
Open the full pipeline board and pick a checkpoint. Before anything happens, Commandr shows which files change back, which new files are removed and which deleted files return.
Chat messages written since that checkpoint are removed too, so the pipeline sees those stages as not run — unless you choose to keep them. A checkpoint of the current state is saved first, so a restore can be undone as well.
What’s never stored
.env files, the paths you protect in Project settings → Folder, and node_modules are never saved in a checkpoint or changed by a restore.
If a checkpoint can’t be saved
Agents that write files stop before they start, and the pipeline pauses with Recovery unavailable and the reason. Fix the disk or repository problem and try again.
You can turn off Require recovery checkpoints in Project settings → Folder, but changes made without a checkpoint may not be recoverable. A warning stays on screen until you turn it back on.
How long they’re kept
The latest 60 checkpoints in each project are kept. They live alongside your repository without being part of it, so git push never sends them.