v5.5.0 · jev judgment points · architecture-survey, diagram & intent skills · sharper deep-interview →

Stop coding alone. Ship with a team of agents.

oh-my-claudecode is the zero-config orchestration layer that turns Claude Code into a coordinated team of 19 specialists — architect, executor, reviewer, and more.

Ship features 3-5× faster. Save 30-50% on tokens. Never compromise on quality.

Install in 30 seconds Star on GitHub 39k ★
~/your-project — claude live
$
↳ autopilot detected. Spinning up a team…
✓ architect planned · 2 files
✓ executor built · 14 files, 420 lines
✓ test-engineer shipped · 36 tests, 100% pass
✓ security-reviewer approved
Done in 4m 38s. 3.2× faster than vanilla Claude Code.

Trusted by 39,000+ developers worldwide

39k
GitHub Stars
19
Specialist Agents
40+
Built-in Skills
100+
Community Articles
● The Problem

Claude Code is powerful. Using it alone is not.

A single agent juggling every role — planning, coding, testing, reviewing — means slower output, more bugs, and wasted tokens on rework. You're leaving 80% of Claude's potential on the table.

BEFORE Solo Claude Code
  • ✕ Sequential tasks — you wait while it thinks, builds, tests, reviews
  • ✕ Opus burns on trivial edits — tokens wasted on the wrong model
  • ✕ Gives up or loops on hard bugs — no verification gate
  • ✕ You remember which command does what, when to use which agent
WITH OMC Orchestrated team
  • ✓ Up to 5 agents in parallel — architect designs while executor builds
  • ✓ Smart model routing — Haiku for lookups, Opus for reasoning
  • ✓ Ralph mode never quits — architect-verified before done
  • ✓ Just describe it in plain English. OMC picks the right mode.
● The Solution

One install. Seven superpowers.

Everything Claude Code should have shipped with. Batteries included, zero config.

◉ Core

19 specialists that know their jobs.

Architect plans. Executor builds. QA tests. Security reviews. Designer polishes. OMC delegates to the right agent automatically — no prompting hacks required.

architect executor qa-tester code-reviewer designer explore security-reviewer + 12 more
3-5×

Faster shipping

Up to 5 workers in parallel. Fullstack features land in minutes.

30-50%

Lower costs

Smart routing sends trivial tasks to Haiku. Opus only when needed.

0

Config lines needed

Install, run setup. Describe what you want. OMC picks the mode.

Canonical in v5

Teams-first architecture

Spin up coordinated squads that share task lists, message each other, and persist until done. No more serial loops.

Ralph mode

Never says "done" until it's verified

Architect-backed verification loops. Security review runs before commit. Pass/fail gates eliminate the "it mostly works" trap.

● Execution Modes

Four modes. One for every mission.

Pick manually, or just describe the work — OMC auto-selects.

AUTOPILOT Default mode

Full autonomy, end to end.

Describe the feature. Autopilot plans with architect, builds with executor, tests with qa-tester, and reviews with code-reviewer — all automatically. The flagship experience.

$ /autopilot "build a todo app with auth"
Sequential
Stage gates
Verified
Before done
Hands-off
Walk away
EXECUTE Persistent execution · verify/fix loops · new in v5

From approved plan to verified code.

Hand Execute a task and it carries it all the way through — implement, verify, fix, repeat — with no silent partials. In v5 it replaces ultrawork, ultrapilot, swarm, and pipeline. Reach for /team when you want coordinated parallel workers, and /verify to cycle a quality gate on its own.

$ /execute "refactor auth to OAuth and get the suite green"
Persistent
No silent partials
Verify → fix
Built-in loop
Canonical
v5 workflow
TEAM Coordinated squad

N coordinated agents. Shared task list.

Spawn a team with a shared pool of tasks. Agents claim atomically, hand off via messages, and synthesize results. The canonical multi-agent surface.

$ /team 3:executor "fix all TypeScript errors in /src"
Atomic
Task claims
Messaging
Inter-agent
Synthesis
Lead-guided
RALPH Relentless persistence

Won't stop until architect-verified.

For hairy bugs and high-stakes refactors. Ralph keeps iterating — plan, fix, verify, repeat — until the Architect agent signs off. Maximum persistence.

$ /ralph "refactor auth to JWT with zero downtime"
Looping
Until verified
Architect gate
Sign-off
Stubborn
By design
● Agent Ecosystem

A specialist for every role on your team.

19 named agents. Teams-first architecture. Specify Haiku, Sonnet, or Opus per invocation.

Analysis

architect debugger tracer

Execution

executor code-simplifier git-master

Search & Research

explore document-specialist scientist

Design & Docs

designer writer

Testing & QA

qa-tester test-engineer

Review & Security

code-reviewer security-reviewer

Planning & Product

planner analyst

Verification

verifier critic

Every agent accepts a model parameter. No tier suffixes. Invoke any agent with haiku, sonnet, or opus.

★ Community Love

What developers are saying.

Pulled from the 100+ articles already published about OMC.

"

The only agents swarm orchestration you actually need. It makes multi-agent coding simple — even for the least experienced Claude Code users.

JN
Joe Njenga
Medium · Hands-on review
"

Specialized teammates operating in parallel with shared task lists and inter-agent messaging. The orchestration pattern modern AI coding needs.

AO
Addy Osmani
Google · The Code Agent Orchestra
"

Zero-config orchestration layer delivering 3-5× speedup and 30-50% token savings through 32 specialized agents. Trended #1 on GitHub.

AK
Alex Kerber
100K Stars in 24 Hours analysis
See all 100+ articles
● 30 seconds · 3 commands

From zero to orchestrated.

Already have Claude Code? You're 30 seconds from a full AI team.

1

Install the plugin

Add OMC to Claude Code's plugin marketplace.

terminal 
$ /plugin marketplace add https://github.com/Yeachan-Heo/oh-my-claudecode
$ /plugin install oh-my-claudecode
2

Run setup wizard

One command. No config files to edit.

terminal 
$ /omc-setup
✓ 32 agents registered
✓ 40+ skills loaded
✓ Configuration complete.
3

Describe the work.

Natural language in. Shipped feature out.

terminal 
$ autopilot build a REST API for managing tasks
↳ Activating autopilot…
  architect · executor · qa-tester · security-reviewer
That's it. Everything else is automatic.
● Recommended · Companion terminal

Now give your agents a terminal of their own.

OMC runs the team. Agentty is where you watch it work — every agent in one window, and you always know which one needs you.

Agentty logo

Agentty

v0.1.14 macOS · Windows · Linux · free

The real terminal for the AI Native era.

Run every AI agent in one window, see what each is doing, and share context between them. A ~17MB Rust-native terminal built for Claude Code, Codex, Gemini CLI and more — the tmux alternative that fits OMC's team workflows.

Unlimited split panes
Claude Code, Codex, Gemini CLI side by side — as many as you want.
Knows what agents are doing
Working, done, or waiting on you — visible at a glance.
Shared context
Search sessions and hand context between agents.
Download Agentty Releases on GitHub

An independent project by Yong Lee (empty-user77). Source to be open-sourced.

● Community & Learning

100+ articles. Start here.

Curated hands-on reviews, deep dives, and tutorials from the OMC community.

marc0.dev May 2026

Claude Opus 4.8 /ultracode: I Ran a 200-Agent Swarm

Hands-on report on Opus 4.8 dynamic workflows — spawning a 200-agent swarm to parallelize research and lift evaluation scores.

By Marco Patzelt →
Medium Jan 2026

I Tested Oh My Claude Code — The Only Agents Swarm Orchestration You Need

Hands-on review of multi-agent orchestration capabilities with real-world examples.

By Joe Njenga →
ROBOCO 🇰🇷 한국어

Oh My Claude Code — Claude Code를 '팀'으로 쓰는 플러그인

In-depth Korean analysis of the plugin's architecture and team-like agent collaboration.

Technical Deep Dive →
Medium Jan 2026

AI Coding Tools: The Complete Guide to Claude Code & Modern Development

Comprehensive guide covering Claude Code ecosystem including plugins like OMC.

By Recep Şen →
The Unwind AI 2025

Claude Code's Hidden Multi-Agent Orchestration — Now Open Source

Deep dive into Claude Code's native agent coordination and how OMC unlocks it.

Technical Analysis →
Medium Feb 2026

What I Learned Building a Trilogy of Claude Code Plugins

First-hand lessons — architecture, best practices, and common pitfalls.

By Pierce Lamb →
AIBit 2026

oh-my-claudecode: Boost Claude Code with Multi-Agent Automation

How OMC transforms Claude Code into a multi-agent system with five execution modes.

AIBit Blog →
ROBOCO 🇰🇷 한국어

Everything Claude Code vs Oh My ClaudeCode — 팀/기업 도입 관점 비교

Enterprise adoption comparison between vanilla Claude Code and OMC plugin ecosystem.

Enterprise Analysis →
AddyOsmani.com Feb 2026

Claude Code Swarms

How agent teams enable specialized teammates to operate in parallel with shared task lists.

By Addy Osmani →
● Official

The source of truth

Full reference, changelog, release history, and the v4 → v5 migration table.

Reference Docs Changelog Releases v4 → v5 Migration
? FAQ

Questions, answered.

Everything you need to know before installing.

What is oh-my-claudecode?
oh-my-claudecode (OMC) is a teams-first multi-agent orchestration layer for Claude Code. It provides 19 specialized agents, 40+ skills, and intelligent workflows that automatically parallelize tasks, optimize token usage, and persist until completion.
Do I need to learn special commands?
No. OMC has zero learning curve. Just describe what you want in natural language, and it automatically detects the best mode and agents to use. Magic keywords like autopilot and ralph are optional shortcuts for power users.
What are the requirements?
Claude Code CLI installed, plus either a Claude Max/Pro subscription or an Anthropic API key. That's it — no other dependencies.
How does token optimization actually work?
OMC uses smart model routing — simple tasks (file reads, syntax fixes) go to Haiku (fast, cheap), complex reasoning (architecture, debugging) goes to Opus. This automatic routing typically saves 30-50% on tokens while maintaining quality.
Autopilot vs Execute vs Team vs Ralph — how do I choose?
Autopilot — default, hands-off, one lead agent from idea to working code (/autopilot "..."). Execute — carries an approved task through implement → verify → fix loops until it's truly done; the v5 replacement for ultrawork, ultrapilot, swarm, and pipeline (/execute "..."). Team — N coordinated agents on a shared task list (/team 3:executor "..."). Ralph — relentless, won't quit until architect-verified. Or just describe the work — OMC picks the right mode.
Is it safe to use with production code?
Yes. OMC ships a dedicated security-reviewer agent, mandatory Architect verification before completion, and verification-before-done protocols. Ralph mode never marks a task complete until it actually works.
How much does it cost?
OMC is 100% free and open source (MIT licensed). You only pay for your Claude Code usage — and thanks to smart model routing, you'll typically spend 30-50% less than vanilla Claude Code.

Ready to ship 3× faster?

Join 39,000+ developers who turned Claude Code into a coordinated team of specialists. Install in 30 seconds. Free forever.

Install oh-my-claudecode Star on GitHub
MIT License Free forever No config