jwangkun/oh-my-claudecode

oh-my-claudecode is a plugin for Claude Code that lets you coordinate multiple AI agents to tackle software projects together. Instead of working with a single assistant, you can spin up a team of agents that plan, write code, verify results, and fix problems in a loop until the job is done. The promise is that you don't need to learn Claude Code's intricacies, you just describe what you want and let it handle the orchestration. At its core, the project runs a staged pipeline: planning, writing a product spec, executing the work, verifying it, and fixing anything that's broken. You can trigger this with simple commands like /team 3:executor "fix all TypeScript errors", which assigns three agents to the task. Beyond its own Claude-based agents, it can also launch separate AI tools, Codex and Gemini, in their own terminal panes so different models work in parallel on different parts of a problem. A built-in "deep interview" mode asks you Socratic questions to clarify vague ideas before any code is written. This is aimed at developers, founders, or technical PMs who use Claude Code but want more power than a single agent provides. If you want to say "build a REST API for managing tasks" and have a coordinated team handle it end to end, this is for you. It also learns from your sessions: when you solve a tricky debugging problem, it extracts that knowledge into a reusable "skill" file that automatically applies next time a similar issue comes up. A few things stand out. It routes simple tasks to faster, cheaper models and reserves expensive models for complex reasoning, which the README claims saves 30-50% on token costs. It also integrates with Telegram, Discord, and Slack for notifications, and can auto-resume sessions when you hit rate limits. The README is dense with configuration options and command references, suggesting it's built for users who want fine-grained control over how their AI agents collaborate.