Oh My ClaudeCode
Team-first multi-agent orchestration for Claude Code, from planning through verification and repair.
Per-dimension scores and reasoning
Evidence shows security configuration (SECURITY.md) and strict mode, but no specific implementation of user confirmation mechanism. Data flow transparency is documented, sensitive data handling has environment variable warnings. Dependency security has version pinning but no vulnerability scan evidence. External effects are documented, rollback mechanism is mentioned but not detailed. Source attribution has author information.
Self-consistency is good, documentation aligns with code. Dependency availability has package.json and CI tests. Failure messages are covered by tests.
Target audience is clear, scenarios are diverse. Capability boundaries are documented. Trigger words are precise. Environment fit has cross-platform tests.
Information architecture is clear, install notes are detailed. Naming is stable. Examples and FAQ are provided. Known limitations are documented. License is clear. Versioning has CI checks. Maintenance responsibility has author and contributing guide.
Output usability has documentation and examples. Marginal value is high, providing multi-agent orchestration. Cost-benefit has cost optimization notes.
Claims are traceable, supported by documentation and code. Cross-source verification has tests. Fact and inference separation is good.
- Publisher identity is unverified, use with caution.
- Dependency security lacks vulnerability scan evidence, consider checking dependency versions.
- User confirmation mechanism is not explicitly implemented, be aware of auto-approval risks.
What does this agent do, and when should you use it?
Oh My ClaudeCode (OMC) is a Claude Code plugin and command-line runtime for multi-agent development workflows. Its in-session interfaces include /team, /autopilot, /ralph, /ultrawork, and /deep-interview, while its terminal interfaces include omc setup, omc ask, and omc team. The /team workflow follows team-plan → team-prd → team-exec → team-verify → team-fix, whereas omc team launches selected provider CLIs in tmux panes on demand. OMC stores runtime state, session data, plans, logs, handoffs, and local artifacts under .omc/ by default, with HUD, session summaries, and replay logs for inspection. It fits developers already working in Claude Code who want coordinated planning, execution, verification, and repair rather than a single prompt-response loop.
After installation, users run /setup or /omc-setup inside a Claude Code session, or omc setup from a terminal. /team sends work through the native staged team workflow; /autopilot, /ralph, and /ultrawork provide other in-session execution modes. omc ask <provider> <prompt> invokes a local Claude, Codex, Gemini, Antigravity, Grok, or Cursor CLI and writes a Markdown artifact under .omc/artifacts/ask/. omc team N:<provider> "..." launches the requested CLI workers in the active tmux session; omc team status and omc team shutdown inspect or stop a named team.
- A Claude Code engineering team fixing an authentication flow can use /team to coordinate planning, implementation, verification, and follow-up fixes.
- A developer who wants independent Codex CLI security review can run omc team 2:codex "review auth module for security issues".
- A solo Claude Code user building an end-to-end feature with minimal ceremony can start /autopilot.
- A product-minded developer with an underspecified idea can use /deep-interview before implementation to clarify requirements.
- An engineering team running headless automation can use deterministic commands such as omc ask and omc session search instead of interactive slash commands.
What are this agent's strengths and limitations?
- Provides both an in-session Claude Code team workflow and terminal-launched tmux workers, covering interactive and shell-based execution surfaces.
- Its documented Team pipeline explicitly separates planning, PRD work, execution, verification, and repair loops.
- omc ask can invoke several local provider CLIs and persists advisory output as Markdown artifacts under .omc/artifacts/ask/.
- Local .omc/ state, HUD output, session summaries, replay logs, and friction reports give the workflow inspectable operational artifacts.
- The core product depends on Claude Code; external Codex and other CLI workers do not make it a standalone Codex-native replacement.
- omc team requires an active tmux session, and each selected provider requires its own installed and authenticated CLI.
- Named Autopilot stage profiles require Linux and flock; unsupported environments reject explicit --workflow invocations.
- Interactive slash commands are not appropriate for CI or headless automation; the documented alternative is deterministic terminal commands.
How do you install or deploy this agent?
Prerequisites: Claude Code CLI plus a Claude Max/Pro subscription or an Anthropic API key. For the recommended plugin route, run these one at a time inside Claude Code:
/plugin marketplace add https://github.com/Yeachan-Heo/oh-my-claudecode
/plugin install oh-my-claudecodeThen run:
/setupAlternatively, install the npm runtime:
npm i -g oh-my-claude-sisyphus@latestThen run:
omc setupInstall tmux when using omc team or rate-limit detection.
How do you use this agent?
For a first task inside Claude Code, run:
/autopilot "build a REST API for managing tasks"For native teams, enable CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 in ~/.claude/settings.json, then run:
/team 3:executor "fix all TypeScript errors"For tmux CLI workers, install and authenticate the selected CLI, enter an active tmux session, then run:
omc team 2:codex "review auth module for security issues"For a provider advisory artifact, run:
omc ask codex "review this patch"How does this agent compare with similar options?
The repository points Codex CLI users to oh-my-codex for the corresponding orchestration experience, while Oh My ClaudeCode is centered on Claude Code. It names oh-my-opencode as an inspiration, but does not provide an auditable feature-by-feature comparison.