Looking for five early TreeHarbor pilots — including failed installs

[ArtAndrew]

I'm looking for the first five people to try TreeHarbor on a disposable repository and tell me where it gets in the way.

TreeHarbor is for agent-platform and developer-infrastructure work: start tasks from an exact revision, keep submitted changes after a worker is lost or a workspace is removed, and return ordinary GitHub PRs. GitHub still owns review, CI and merge. The core is Rust; source is available under ELv2 with no workspace-count limit.

Try 0.1.19-alpha.1. Installation guide · Signed Linux packages.

See what you get first. In the Linux 0.1.16 trial, two scripted agents edited the same file from one commit. Their changes became draft PR #8 and draft PR #9. Both patches still exported to Git after workspace cleanup. The 36-second walkthrough illustrates an earlier run.

For your first trial, use a disposable GitHub repository. Setup needs Linux, PostgreSQL and a GitHub App for PR publication. The managed walkthrough uses Docker and gVisor. The worktree comparison shows timings for both the first and repeat batches.

A useful report can be short:

• What task are you trying to automate, and what do you use today?
• How far did you get? What was the first blocker?
• How much time went into infrastructure setup versus the TreeHarbor task?
• Would you try a second task? If not, why?

Open a pilot report, or use Q&A before installing. Failed attempts and “Git worktrees already solve this for me” are useful answers too.

Please keep private source, credentials and customer details out of public posts. No code contribution is required; external code PRs still need the contribution-rights process described in CONTRIBUTING.