Repository navigation
Using jj workspaces for parallel coding agents #10442
Replies: 1 comment
|
Thanks for writing this up and asking how others do it. Disclosure: I build Carbon Forge at 100monkeys.ai, which takes a different route to the same problem. Agents don't get a working copy at all. Each agent opens a lane, a set of patches over the trunk, and every read on the lane composes the trunk as it stands now with those patches, so there is no workspace to rebase. It never merges text by itself: a patch that no longer applies comes back as a named conflict, the agent resolves it against the trunk it saw, and a landing is one changeset, written whole or not at all. Git stays a projection pushed to GitHub fast-forward only. The trade-off is real: there is no offline work and no local checkout, which jj's model doesn't ask you to give up. The design notes are here if the comparison is useful: https://carbonforge.dev/docs/p/features/source-control-on-the-substrate?utm_source=github&utm_medium=social&utm_campaign=push-2026-10-28 |
Uh oh!
There was an error while loading. Please reload this page.
I've been using jj workspaces to run several Claude Code agents on the same repo.
Wrote up my workflow for inspecting changes, rebasing multiple workspaces, and recovering lost edits. Also covered a few limitations I've run into. Would be interested to hear how others approach this.
https://minikin.me/blog/jj-workspaces-for-parallel-agents/
All reactions