Multi-Agent Orchestration
21K+ StarsTypeScriptby Yeachan Heo · Team Mode + Deep Interview + Autopilot
oh-my-claudecode transforms Claude Code from a single assistant into a managed development team. Its core innovation is Team Mode — a five-stage pipeline where separate agents handle planning, requirements, implementation, verification, and remediation as distinct, accountable steps.
The project also introduces Deep Interview, a Socratic requirements-mining phase that runs before any code is written, and Autopilot mode for hands-off execution of well-defined tasks. With 21K+ GitHub stars and multi-language documentation, it has become one of the most adopted Claude Code enhancement tools.
team-plan
team-prd
team-exec
team-verify
team-fix
Pipeline Architecture
Creates the implementation strategy by analyzing the task, identifying components, estimating complexity, and defining the execution order. The plan serves as the contract between subsequent stages.
Stage Outputs
Generates a detailed product requirements document from the plan. Specifies acceptance criteria, edge cases, API contracts, and test expectations. This document becomes the verification target.
Stage Outputs
Implements the code according to the PRD. The executing agent works within the boundaries defined by the plan and PRD, reducing the risk of scope creep or architectural drift during implementation.
Stage Outputs
Validates the implementation against the PRD acceptance criteria. Checks functional correctness, edge case handling, test coverage, and code quality. Produces a pass/fail report with specific findings.
Stage Outputs
Addresses issues identified during verification. Fixes are targeted at specific findings rather than general code improvement, keeping changes focused and reviewable. The cycle can loop back to team-verify.
Stage Outputs
Requirements Mining
Deep Interview is oh-my-claudecode's answer to the most expensive problem in software development: building the wrong thing. Before any code is written, the system asks a structured series of questions designed to surface ambiguity, missing requirements, edge cases, and implicit assumptions.
The interview output feeds directly into the team-plan and team-prd stages, ensuring that the implementation targets are grounded in explicitly validated requirements rather than the developer's initial — often incomplete — description.
What It Asks
Scope boundaries, user types, error scenarios, performance requirements, integration points, and acceptance criteria.
When to Use It
For any task where requirements are not fully specified, ambiguous, or where the cost of rework is high.
When to Skip It
For well-defined, small-scope tasks where requirements are already clear. Use Autopilot mode instead.
Comparison
Three approaches to Claude Code mastery. oh-my-claudecode orchestrates teams. Superpowers enforces methodology. Best Practice documents the patterns behind both.
| Dimension | oh-my-claudecode | Superpowers | Best Practice |
|---|---|---|---|
| Core model | Team pipeline orchestration | Skills-driven methodology | Knowledge documentation |
| Agent count | Multiple (staged) | Subagent-driven | N/A (reference) |
| Requirements phase | Deep Interview + PRD | Spec-first questioning | Documented patterns |
| Verification | team-verify stage | Automatic inspection | Testing best practices |
| Install method | npm / marketplace | Marketplace / CLI | Read on GitHub |
| Stars | 21K+ | 131K+ | 30K+ |
Execution Brief
Use this page as a rollout checklist, not just reference text.
Tool Mapping Lens
Catalog-oriented pages work best when users can map discovery, evaluation, and rollout in a clear path instead of reading an undifferentiated list.
Use this board for oh-my-claudecode before rollout. Capture inputs, apply one decision rule, execute the checklist, and log outcome.
Input: Objective
Deliver one measurable improvement with oh-my-claudecode multi-agent orchestration
Input: Baseline Window
20-30 minutes
Input: Fallback Window
8-12 minutes
| Decision Trigger | Action | Expected Output |
|---|---|---|
| Input: one workflow objective and release owner are defined | Run preview execution with fixed acceptance criteria. | Go or hold decision backed by repeatable evidence. |
| Input: output quality below baseline or retries increase | Limit scope, isolate root issue, and rerun controlled test. | One confirmed correction path before wider rollout. |
| Input: checks pass for two consecutive replay windows | Promote to broader traffic with fallback path active. | Stable rollout with low operational surprise. |
tool=oh-my-claudecode multi-agent orchestration objective= preview_result=pass|fail primary_metric= next_step=rollout|patch|hold
oh-my-claudecode is a multi-agent orchestration framework that transforms Claude Code from a single AI assistant into a structured development team. Created by Yeachan Heo, it introduces Team Mode — a five-stage pipeline where separate agents handle planning, product requirements, implementation, verification, and remediation as distinct accountable steps.
The framework addresses a fundamental limitation of single-agent development: when one agent handles everything from requirements analysis to code writing to testing, the quality of each phase suffers because there is no separation of concerns. oh-my-claudecode enforces separation by assigning each phase to a dedicated agent with a specific role and output format.
Beyond Team Mode, oh-my-claudecode includes Deep Interview for Socratic requirements mining and Autopilot mode for hands-off execution of well-defined tasks. The project supports multiple languages (English, Korean, Chinese, Japanese, Spanish, Vietnamese, Portuguese) and is available via npm and the Claude Code marketplace.
With over 21,000 GitHub stars and active development, oh-my-claudecode has established itself as the leading multi-agent orchestration solution for Claude Code. It is particularly popular with developers who need production-quality output from Claude Code without manually managing the quality control process.
Install oh-my-claudecode from the Claude Code marketplace or via npm (oh-my-claude-sisyphus). After installation, Team Mode commands become available in your Claude Code session. Start with a simple feature to learn the pipeline before applying it to complex tasks.
Use Deep Interview for any task where you are not fully confident in the requirements. The interview process takes a few minutes but saves significant rework by surfacing missing requirements, edge cases, and implicit assumptions before any code is written.
For well-defined, straightforward tasks, use Autopilot mode instead of the full Team Mode pipeline. Autopilot handles the entire cycle automatically and is best suited for tasks like "build a REST API for X" or "refactor Y to use Z pattern" where the scope and approach are clear.
Combine oh-my-claudecode with Superpowers skills for maximum effect. Superpowers enforces methodology (spec-first, TDD) within each pipeline stage, while oh-my-claudecode ensures the stages themselves happen in the right order with proper handoffs. The combination produces consistently high-quality output.
Treat this page as a decision map. Build a shortlist fast, then run a focused second pass for security, ownership, and operational fit.
When a team keeps one shared selection rubric, tool adoption speeds up because evaluators stop debating criteria every time a new option appears.
Outcome: A production-ready REST API with complete test coverage, proper error handling, and verified acceptance criteria — delivered in a single structured session.
Outcome: A clean refactor completed hands-off, with verified test preservation and no manual quality control needed.
Outcome: Deep Interview prevented building a simple notification component when the actual requirement was a multi-channel notification service — saving days of rework.
oh-my-claudecode is a multi-agent orchestration framework for Claude Code created by Yeachan Heo. It introduces Team Mode — a staged pipeline where multiple Claude Code agents work through a structured plan-PRD-execute-verify-fix loop. Instead of treating Claude Code as a single assistant, oh-my-claudecode organizes it as a managed development team with specialized roles.
Team Mode runs a five-stage pipeline: team-plan creates the implementation strategy, team-prd generates a detailed product requirements document, team-exec implements the code, team-verify validates the implementation against the PRD, and team-fix addresses any issues found during verification. Each stage produces artifacts that feed into the next, creating a structured development cycle.
Deep Interview is a Socratic questioning system that runs before any code is written. It asks clarifying questions about requirements, constraints, edge cases, and acceptance criteria. The goal is to surface ambiguity and missing information upfront — before it becomes expensive rework during implementation. The interview output feeds directly into the team-plan and team-prd stages.
Autopilot mode is a single-command execution path for well-defined tasks. Instead of running the full team pipeline, you describe what you want (e.g., "autopilot: build a REST API for managing tasks") and oh-my-claudecode handles the entire cycle automatically. It is best suited for tasks where requirements are already clear and the developer wants hands-off execution.
oh-my-claudecode can be installed via the Claude Code plugin marketplace or through npm (package name: oh-my-claude-sisyphus). There is also a Codex variant (oh-my-codex) for OpenAI Codex users. After installation, Team Mode and Autopilot commands become available in your Claude Code session.
Superpowers is a methodology framework that enforces spec-first, TDD-driven development through skills that activate automatically. oh-my-claudecode is a team orchestration framework that coordinates multiple agents through staged pipelines. Superpowers focuses on how a single agent should behave. oh-my-claudecode focuses on how multiple agents should collaborate. They can be used together.
Yes. Despite the "team" branding, oh-my-claudecode is designed for solo developers who want the quality benefits of team-based development processes. The staged pipeline ensures that planning, implementation, and verification happen as distinct steps rather than being mixed together in a single conversation — a common source of quality issues in unstructured Claude Code usage.
Send the exact workflow you are solving and we will prioritize a new comparison or rollout guide.