graft

Personal plan

Not available yet. Join the waitlist.

Your own copy, kept on upstream's releases

For one person, or their agent, with a changed copy of an open-source project. We watch upstream, carry your changes onto each new release in a sandbox, and open a pull request when the result passes your checks. Hosted only. No one reviews it, so a strict rule does that job instead.

It never moves you onto a copy that fails your checks.

That's the promise, and it's narrower than "it's always right" on purpose. Agents miss some calls and tests don't catch everything. So the rule is built to stay put when in doubt:

  • Move only when everything lands. Every change is kept, dropped because the new release already has it, or rewritten. If any change is stuck, nothing moves.
  • Move only when the checks pass. Your build and tests run on the carried copy and on the new release as shipped. The carry passes when it adds no failure upstream doesn't have.
  • No unchecked rewrites. An agent's rewrite is accepted only when tests exist to check it, the agent is at least 0.8 sure, and agents agree.
  • Otherwise, stay. Your copy stays where it was, your old branch is never touched, and the report says what stopped it.
The safety rule, two syncs
Next releasesync 1
every change lands
checks pass
agents sure, agree
Movedpull request opened
Newest releasesync 2
every change lands
checks pass
Stayedreport says why
Real forks on GitHub, agents and tests on10 of 24 moved, 0 wrong
Debian release history, no agents5 of 10 moved, 0 of 62 wrong
Staying is the safe failure. A sync that can't prove itself leaves your copy alone.

What we've measured so far

We ran the sync on real forks and on Debian's release history before offering anything. These are the results as of 2026-10-28, and we update this page when we rerun them.

10 of 24syncs on 12 real forks on GitHub moved by themselves, with agents and the project's own tests
0wrong moves among them; every moved result was read by hand
5 of 10Debian curl and openssh releases moved by themselves, with no agents
0 of 62changes in those moves were wrong

The forks were personal, small-team and company forks: 7 of 12 syncs onto the next release moved, and 3 of 12 onto the newest. Most syncs that stayed did so because an agent wasn't sure enough, or because the fork's own changes already failed upstream's tests. Syncing each release as it ships moves more often than catching up later. The data and method are in the evidence repository.

Plans and timing

  • StatusPlanned. Not available to buy or use yet.
  • How it runsHosted only. We run the sync in a sandbox and open a pull request on your copy with the report.
  • PriceA low monthly price per copy, model calls included. We'll set it before launch and tell the waitlist first.
  • Before launchAutomatic test setup for each project. Getting a project's tests to run was the hardest part of the benchmark.

Join the waitlist

One email when it opens. Your address is used for nothing else. See privacy.