oh-my-claudecode — Zero-Config Multi-Agent System

What it is

oh-my-claudecode is a multi-agent orchestration system for Claude Code that requires zero configuration. It provides team mode for collaborative agent work, autopilot for hands-off execution, Ralph persistent execution for long-running tasks, and ultrawork parallel mode for running multiple agents simultaneously. The system includes 19 specialized agents covering different development roles.

This tool is for Claude Code power users who want to scale their AI-assisted development beyond a single agent. Instead of one Claude instance handling everything, you get a coordinated team of specialized agents.

How it saves time or tokens

By splitting work across specialized agents, oh-my-claudecode reduces the context window load on any single agent. Each agent focuses on its domain: one reviews code, another writes tests, another handles documentation. Parallel mode runs multiple agents at once, completing tasks that would take hours in sequence in minutes. The estimated token cost is around 3,500 tokens per session.

How to use

1. Install oh-my-claudecode in your Claude Code environment.
2. Activate team mode or autopilot.
3. Assign tasks and let the agents coordinate.
4. Review results from the orchestrated workflow.

# Install oh-my-claudecode
git clone https://github.com/anthropics/oh-my-claudecode.git
cd oh-my-claudecode
./install.sh

# Activate team mode
claude /team 'Build a REST API for user management'

# Run in autopilot mode
claude /autopilot 'Refactor auth module and add tests'

# Use ultrawork parallel mode
claude /ultrawork 'Review all TypeScript files in src/'

Example

Team mode in action:

[Architect] Designing API schema for user management...
  -> 4 endpoints: GET /users, POST /users, PUT /users/:id, DELETE /users/:id

[Coder] Implementing routes and handlers...
  -> Created src/routes/users.ts with validation

[Tester] Writing unit and integration tests...
  -> Created tests/users.test.ts with 12 test cases

[Reviewer] Reviewing implementation...
  -> Found 2 issues: missing error handling in DELETE, no pagination on GET

[Coder] Fixing issues from review...
  -> Added error handling and pagination support

[Done] All agents completed. 5 files created, 0 issues remaining.

Related on TokRepo

• Multi-agent frameworks — Compare multi-agent approaches
• AI coding tools — More AI development tools

Common pitfalls

• Running many agents in parallel consumes tokens quickly. Monitor your API usage when using ultrawork mode.
• Agent coordination adds overhead. For simple tasks, a single Claude Code session is faster and cheaper.
• The 19 specialized agents may not all be relevant to your workflow. Identify which agents provide value for your use case.
• Autopilot mode makes decisions without asking. Review its output carefully, especially for destructive operations like file deletion.
• oh-my-claudecode depends on Claude Code's architecture. Updates to Claude Code may require oh-my-claudecode updates.