Shep AI
Shep AI is an open-source, local-first orchestrator for running fleets of AI coding agents in parallel. It gives each feature an isolated Git worktree and drives work from a one-line description through optional specification gates, implementation, commit, push, CI monitoring/auto-fix, and a draft pull request.
At a Glance
- Individual software engineers
- AI-assisted developers and technical founders
- Engineering teams using GitHub pull requests and CI
- Developers working on monorepos or multiple repositories
- +1 more
AI Tools by Shep AI
(1)
Shep
Parallel AI Coding Agent Orchestrator
Discussions
No discussions yet
Be the first to start a discussion about Shep AI
Latest News
Shep v1.242.0: chat-first web page with inline specification, preview, and plan blocks.
Shep v1.241.0: decision.answered and decision.defaulted telemetry, Git checkpoints after each feature-agent step, and SDLC board feature-flag cleanup.
Shep v1.240.0: review cleanup with per-area feature flags and ASPM off by default.
Shep v1.239.0: added opt-out usage metrics.
Products & Services
An MIT-licensed command-line orchestrator installed with npx or npm. It creates isolated worktrees and branches, runs configured coding agents, optionally applies spec/approval gates, commits and pushes changes, opens draft pull requests, watches CI, and retries fixes.
A local web control center, normally opened at localhost:4050, with a live graph of repositories, features, agent runs, status, diffs, chat, and IDE handoff. app.shep.bot provides a browser-based entry point.
An experimental query-aware agent harness released as part of the v1.235.0 release.
Market Position
Shep positions itself as the orchestration and workflow layer around coding agents rather than as a replacement model or coding agent. Its differentiators are parallel execution, worktree isolation, agent/model portability, optional spec gates, CI auto-fix, and user-controlled review/merge. The repository’s competitor documentation references T3 Code; Shep also overlaps with direct use of individual coding-agent CLIs and other AI software-development orchestration tools.
Founding Story
The project was started to make parallel AI-assisted software development practical without branch conflicts or losing developer control. Its initial vision was an agent-agnostic, local-first layer that manages worktrees, the software-development lifecycle, CI feedback, and pull requests while leaving review and merge decisions with the user.
Business Model
Revenue Model
The software is free, open source, and MIT licensed. It is distributed through the @shepai/cli npm package and can be run with npx; users bring their existing agent subscriptions or API keys. The site also presents app.shep.bot as free.
Pricing Tiers
MIT-licensed software; local CLI and dashboard. Users supply the subscription or API credentials for the coding agent they choose.
Target Markets
- Individual software engineers
- AI-assisted developers and technical founders
- Engineering teams using GitHub pull requests and CI
- Developers working on monorepos or multiple repositories
- Organizations seeking local-first, agent-agnostic developer tooling
- Running multiple independent software features concurrently
- Shipping AI-generated changes through a normal branch, CI, review, and pull-request process
- Managing complex work with requirements and plan approval gates
- Starting a new application and incrementally adding features
- Developing across multiple repositories while keeping the main checkout untouched
- Using different coding-agent CLIs per feature without vendor lock-in