ClipboardHealth

ClipboardHealth / groundcrew

Public

Farm out ready tickets to sandboxed coding agent CLIs in git worktrees.

17
1
89% credibility
Found May 18, 2026 at 21 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

Groundcrew is an AI agent orchestrator that watches a Linear project and automatically assigns labeled tickets to AI coding agents like Claude or Codex. It creates isolated git workspaces for each ticket, manages parallel agent execution, and automatically cleans up workspaces when tickets are marked complete. The tool supports both local execution (on macOS with Safehouse security) and remote execution (via a configured remote runner), making it suitable for teams wanting to automate AI-assisted development workflows.

How It Works

1
📋 Create tickets in Linear

You write tickets in your Linear project and add them to a dedicated project folder.

2
🏷️ Label tickets for automation

You add an 'agent-claude' or 'agent-codex' label to tickets you want AI to handle.

3
🚀 Groundcrew starts watching

The tool continuously monitors your project and automatically picks up new labeled tickets.

4
Where should the agent work?
🍎
Local on Mac

Agents run in isolated workspaces on your machine with Safehouse security

☁️
Remote server

Agents run on a configured remote runner (useful for Linux or heavy workloads)

5
🛠️ Workspace creation

Groundcrew creates a fresh isolated workspace for each ticket using git worktrees.

6
🤖 AI agent gets to work

The AI agent receives the ticket details and starts coding in its own workspace.

✅ Ticket completed and cleaned up

When you mark the ticket done, Groundcrew automatically closes the workspace and removes the worktree.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 21 to 17 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is groundcrew?

Groundcrew is a TypeScript CLI that watches a Linear project and farms out ready tickets to sandboxed AI coding agents running in git worktrees. It polls for tickets labeled `agent-*`, provisions isolated terminal workspaces, launches the configured agent CLI, and cleans up when tickets reach a terminal status. The `crew` binary handles the full lifecycle: dispatching work, tracking capacity, respecting usage limits, and managing blocker dependencies.

Why is it gaining traction?

The hook is treating AI agents like a shared worker pool. Instead of manually opening tmux panes and babysitting one agent at a time, groundcrew queues tickets, gates them against session limits, and routes work to the agent with the most available capacity. It handles the boring parts: creating worktrees, authenticating against remote runners, forwarding build secrets, and cleaning up stale state. Teams using multiple agents (Claude, Codex) get automatic load balancing via `agent-any` routing.

Who should use this?

Engineering teams running Linear as their project tracker and wanting to scale AI coding agents across many concurrent tickets. It is specifically useful for teams with a backlog of routine coding tasks that can be handed off to agents without constant supervision. Organizations already invested in Claude Code or Codex and wanting to automate ticket-to-agent dispatch will get the most value. macOS users with Safehouse get local execution; Linux/WSL users are limited to remote runner dispatch.

Verdict

Groundcrew solves a real workflow problem for teams scaling AI agents, but the 0.899% credibility score and 17 stars reflect an early-stage, internal tool that ClipboardHealth open-sourced. The documentation is thorough and the test coverage hits 100% thresholds, which is a good sign for maintenance. If you run Linear and want automated ticket dispatch to coding agents, this is worth a pilot. Just budget time for setup: Safehouse, cmux/tmux, codexbar, and Linear API credentials are prerequisites before `crew run --watch` does anything useful.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.