Launching today

cowtree
A separate workspace for every AI agent, at no disk cost
3 followers
A separate workspace for every AI agent, at no disk cost
3 followers
Running several AI coding agents on one project? They overwrite each other's files, or each needs a full copy of your code. cowtree gives every agent its own instant copy that uses almost no disk. Works with Claude Code, Codex and Cursor. Free, open source.





Free
Launch Team / Built With

AWS Startup AdvisorPersonalized AWS guidance that meets you where you’re already building
Promoted

Hi Product Hunt, I'm Paa Kofi, and I built cowtree because disk space became the limit on how many AI coding sessions I could run.
I run several Claude Code and Codex sessions on the same repository, one branch each. Each needs its own working tree, and git worktree add gives you that, but every worktree is a full copy of every file, and node_modules has to be reinstalled in each. On a nearly full laptop that stopped me at two or three sessions.
cowtree builds the worktree out of APFS clonefile copies of the main checkout instead (reflink on btrfs/XFS), ignored build directories included. A 704 MB checkout becomes a 3.5 MB worktree, and editing a file costs that file. No FUSE, no daemon: a few hundred lines of bash and a small Python helper, so editors, compilers and git see ordinary files.
It also records which branch lives in which worktree in git config (cowtree which, cowtree branches), sweeps worktrees whose PRs merged (squash merges too), turns your existing plain worktrees into clones (cowtree adopt), and ships as a Claude Code plugin and an MCP server so agents use it instead of git worktree add.
Install: brew install paakofiaidoo/tap/cowtree
Two things I'd love feedback on: the Linux reflink backend is written but hasn't been exercised on a real btrfs/XFS box yet, and I'd like to know which agent setups (Cursor, Windsurf, Zed, Cline) you'd want first-class support for.