shep-ai · 2026-02-02 · notable
Shep — parallel AI coding agent orchestrator across git worktrees
MIT-licensed CLI and web dashboard that runs multiple AI coding agents in parallel, each in an isolated git worktree. Supports Claude Code, Cursor, Codex, and 5 more; automates the full prompt-to-PR lifecycle with CI watching and auto-fix.
Run ten feature branches in parallel — each with its own AI agent, git worktree, and PR — from a single CLI command.
What is it?
Shep is an open-source control center for running multiple AI coding agents at the same time. Each agent works in an isolated git worktree on its own branch, so concurrent agents never step on each other's files. A web dashboard at localhost:4050 tracks every agent session; a CLI handles the trigger side. It supports eight agents: Claude Code, Cursor CLI, Gemini CLI, Codex, GitHub Copilot, and others.
How does it work?
Running `shep feat new "add stripe payments" --push --pr` creates a git worktree, spins up the chosen agent inside it, watches the agent complete the task, commits the result, pushes the branch, and opens a pull request — all unattended. After the PR opens, Shep watches CI; if a job fails, it feeds the failure logs back to the agent for an auto-fix round. An optional spec-driven mode adds structured stages (Requirements → Research → Plan → Implement) with three human approval gates. All state is stored locally in ~/.shep/ as SQLite.
Why does it matter?
The bottleneck in AI-assisted development is usually context switching: you trigger one agent, wait, then trigger the next. Shep removes that wait by parallelizing across worktrees, which are native to git and carry no merge risk. The CI auto-fix loop means failed builds do not require manual re-prompting.
Who is it for?
Solo devs and small teams using AI coding assistants who want to parallelize feature work.
Try it
npx @shepai/cli