simonrw For a parallel environment in a jj repo, reach for a jj workspace, not a git worktree. A workspace is a second working copy backed by the same repo, so parallel tasks never fight over one checkout. djo (the dojjo tool) manages the workspace lifecycle — create, hydrate, merge, clean up — and runs hooks at each stage.
Requires a jj repo and the djo binary on PATH; see reference.md to install it or set up hooks.
Run every step from inside the repo.
1. Create the workspace
djo switch <name>
Creates a jj workspace plus a bookmark <name>, in a sibling directory ../<name> by default. cd into that directory — djo switch cannot change your shell's directory for you.
Done when djo list shows <name> and your shell is in its directory.
2. Hydrate untracked files from the original
A fresh workspace holds only jj-tracked files. Untracked, git-ignored files — node_modules, .env, build caches — are absent, so the toolchain fails until you bring them over.
djo copy-ignored --from default
default is jj's original workspace; name a different source if you branched from one. Add --dry-run to preview, --force to overwrite. A post-start hook can do this automatically (see reference.md).
Done when the build and test commands run without missing-dependency or missing-config errors.
3. Work in the workspace
Edit and commit with jj as normal. The original working copy and any other workspaces are untouched.
4. Merge back
djo merge <target>
Squashes the workspace's changes, rebases onto <target>, moves the bookmark, and removes the workspace directory. If it fails partway, run jj op undo to revert, then retry. Tune squash/rebase/push behaviour in config (see reference.md).
Done when <target> contains the changes and the workspace directory is gone.
5. Clean up
djo prune
Removes every workspace already merged into the default branch. To drop one workspace you are abandoning unmerged, use djo remove <name>.
Done when djo list shows only the workspaces you still want.