Definition
Oh-My-ClaudeCode (OMC) is a multi-agent orchestration system for Claude Code with a zero learning curve. It stands out for Team mode (a canonical staged pipeline), multi-provider orchestration (Claude + Codex + Gemini), a deep interview system, and persistent execution modes. NPM package: oh-my-claude-sisyphus.
Team mode runs a shared task from plan to PRD, then exec, verify, and fix. One mode runs a tmux team of Claude, Codex, and Gemini workers. Others cover tri-model synthesis, a single lead autopilot, a parallel burst, a persistent verify and fix loop, and staged processing. The deep interview asks questions until the request is clearer.
Basic Information
| Item | Value |
|---|---|
| Stars | 28.4k |
| Forks | 2.6k |
| Top Collaborators | 5+ active |
| License | MIT |
| npm package | oh-my-claude-sisyphus |
Installation
Plugin Marketplace (Recommended)
/plugin marketplace add https://github.com/Yeachan-Heo/oh-my-claudecode
/plugin install oh-my-claudecode
NPM
npm i -g oh-my-claude-sisyphus@latest
Setup
/setup # Inside Claude Code session
omc setup # From terminal
Core Orchestration Modes
| Mode | What it is | Use For |
|---|---|---|
| Team (recommended) | team-plan -> team-prd -> team-exec -> team-verify -> team-fix |
Coordinated agents on shared task |
| omc team (CLI) | tmux CLI workers: claude/codex/gemini in split-panes |
Cross-provider tasks |
| ccg | /ask codex + /ask gemini, Claude synthesizes |
Mixed backend+UI work |
| Autopilot | Autonomous execution (single lead agent) | End-to-end features |
| Ultrawork | Maximum parallelism (non-team) | Burst parallel fixes |
| Ralph | Persistent mode with verify/fix loops | Must-complete tasks |
| Pipeline | Sequential, staged processing | Multi-step transformations |
Key Features
Team Mode (Canonical)
/team 3:executor "fix all TypeScript errors"
Staged pipeline: plan -> PRD -> exec -> verify -> fix (loop)
Enable Claude Code native teams:
{
"env": {
"CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1"
}
}
tmux CLI Workers (v4.4.0+)
omc team 2:codex "review auth module"
omc team 2:gemini "redesign UI components"
omc team 1:claude "implement payment flow"
omc team status auth-review
omc team shutdown auth-review
Workers spawn on demand and die when the task completes.
Deep Interview
/deep-interview "I want to build a task management app"
Socratic questioning to clarify your thinking before coding. Measures clarity across weighted dimensions.
19 Specialized Agents
| Agent Tier | Purpose |
|---|---|
| Architecture | System design decisions |
| Research | Information gathering |
| Design | UI/UX decisions |
| Testing | Quality assurance |
| Data Science | Data analysis |
| Executor | Task implementation |
Smart model routing: Haiku for simple tasks, Opus for complex reasoning.
Custom Skills
# .omc/skills/fix-proxy-crash.md
---
name: Fix Proxy Crash
description: aiohttp proxy crashes on ClientDisconnectedError
triggers: ["proxy", "aiohttp", "disconnected"]
source: extracted
---
Wrap handler at server.py:42 in try/except ClientDisconnectedError...
| Scope | Path | Shared with |
|---|---|---|
| Project | .omc/skills/ |
Team (version-controlled) |
| User | ~/.omc/skills/ |
All projects |
Manage: /skill list | add | remove | edit | search
Auto-learn: /learner extracts patterns
Provider Advisor
omc ask claude "review this migration plan"
omc ask codex "identify architecture risks"
omc ask gemini "propose UI polish ideas"
HUD Statusline
Real-time orchestration metrics in the status bar.
Notifications
Telegram, Discord, and Slack notifications when sessions stop:
omc config-stop-callback telegram --enable --token <bot_token> --chat <chat_id>
omc config-stop-callback discord --enable --webhook <url>
omc config-stop-callback slack --enable --webhook <url>
OpenClaw Integration
Forwards Claude Code session events to the OpenClaw gateway for automated responses.
Autoresearch
omc autoresearch --mission "improve startup performance" --eval "npm test"
Rate Limit Wait
omc wait # Check status
omc wait --start # Auto-resume daemon
In-Session Shortcuts
| Shortcut | Kind | Effect |
|---|---|---|
/team |
Slash skill | Team orchestration |
/ccg |
Slash skill | Tri-model synthesis |
/autopilot |
Skill/trigger | Autonomous execution |
/ralph |
Skill/trigger | Persistence mode |
/ultrawork / ulw |
Skill/trigger | Maximum parallelism |
/ralplan |
Skill/trigger | Iterative planning |
/deep-interview |
Slash skill | Socratic requirements |
deepsearch |
Trigger | Codebase-focused search |
ultrathink |
Trigger | Deep reasoning |
Requirements
| Requirement | Notes |
|---|---|
| Claude Code CLI | Required |
| Claude Max/Pro OR API key | Required |
| tmux | For team mode, rate-limit detection |
| Codex CLI (optional) | For cross-provider orchestration |
| Gemini CLI (optional) | For cross-provider orchestration |
Cost Estimate
3 Pro plans (Claude + Gemini + ChatGPT) = ~$60/month
Tech Stack
| Component | Technology |
|---|---|
| Language | TypeScript |
| Runtime | Node.js |
| Package | npm (oh-my-claude-sisyphus) |
| Process | tmux |
| Notifications | Telegram, Discord, Slack webhooks |
Pros
| Pro | Description |
|---|---|
| Zero learning curve | Natural language, auto-delegation |
| Multi-provider | Claude + Codex + Gemini orchestration |
| Team mode | Canonical staged pipeline |
| Deep interview | Socratic requirements clarification |
| Custom skills | Auto-learn and auto-inject |
| HUD statusline | Real-time metrics |
| 28k+ stars | Large community |
| Persistent execution | Ralph mode keeps going |
| Cost optimization | Smart model routing (30-50% token savings) |
Cons
| Con | Description |
|---|---|
| tmux dependency | Needs tmux for team mode |
| Multiple subscriptions | ~$60/month for full multi-provider |
| Complex architecture | Many modes and commands |
| Package naming confusion | npm: oh-my-claude-sisyphus != repo name |
| Windows support | Needs psmux for native Windows |
| Resource intensive | Multi-agent orchestration uses tokens |
When to Use
- Multi-agent orchestration: Need a coordinated team of agents
- Cross-provider workflows: Want to combine Claude + Codex + Gemini
- Complex features: Need a staged plan -> exec -> verify pipeline
- Requirements clarification: Want a Socratic deep interview before coding
- Persistent execution: Want tasks to complete fully (Ralph mode)
- Large teams: Need shared skills and coordination
The page lists 19 specialized agents in tiers such as architecture, research, design, testing, data science, and executor. Simple tasks go to Haiku and complex reasoning goes to Opus. Project skills live in .omc/skills. User skills live in ~/.omc/skills and apply to every project. A HUD status line shows live orchestration metrics. Telegram, Discord, or Slack can report when a session stops. Cross-provider work needs Claude Code, a Max or Pro plan or an API key, and tmux. Three Pro plans, Claude plus Gemini plus ChatGPT, come to about 60 dollars a month. The npm package is named oh-my-claude-sisyphus, which does not match the repo name.
References:
See also: Kit Comparison · Multi Agent Systems · Everything Claude Code