Back to Ecosystem Pulse
ToolExperimental
shep
by shep-ai
Run parallel AI agent pipelines in isolated git worktrees
TypeScript
Updated Jul 7, 2026
Share:
Summary
Orchestrate Orchestration of parallel AI agents across git worktrees to ship features faster. Uses a CLI-driven workflow that spins agent tasks into isolated worktrees so multiple agent runs, experiments, or proposals can proceed concurrently without blocking the main repo. Distinctive features include worktree-based isolation, scripted agent pipelines, and developer-friendly CLI integration for code-centric agent workflows.
The Value Proposition
As teams use multiple agents to generate, review, and iterate on code, reproducible isolation and side-by-side comparison become crucial for trust. By recording agent outputs in separate git worktrees, Shep makes it practical to compare proposals, audit changes, and track an agent's track record across experiments. That makes it easier to evaluate agent reliability, reproduce failures, and adopt evaluation patterns (A/B agent runs, continuous agent evaluation) before merging outputs into mainline code. This aligns with Swarm Intelligence principles. Swarm Intelligence Evaluation-Driven Development (EDDOps)
Ideal For
Developers and teams building code-centric agent workflows who need isolated, reproducible agent runs and side-by-side comparisons before merging. Market-Based Coordination Pattern
Use Cases
- Run concurrent agent proposals on the same codebase without merge conflicts using isolated worktrees
- Compare outputs from different LLMs or agent strategies side-by-side to decide which to merge
- Reproduce and audit agent-generated PRs and trace agent failure modes across experiments
Works With
gitcliopenaianthropicgoogle-gemini
Topics
agent-orchestrationagentic-aiaiai-agentsai-code-generationai-codingai-nativeautomationclaudeclaude-code+10 more
Similar Tools
autogencrewai
Keywords
multi-agent orchestrationagent delegationagent track record