wtf is claude-forge?

claude-forge is an add-on system for Claude Code, the command line coding assistant, that turns it from one assistant working alone into a coordinated team of specialized agents. Instead of handling a whole task yourself, you type a single command like forge build me a landing page, and the system classifies the request, picks a right sized team from its 18 built-in agents, splits the work into pieces, has different agents build the pieces in parallel, and then runs tests and a review pass before reporting back what actually happened. The project ships with 18 agents covering roles like building, testing, reviewing, security, SEO, and documentation, along with 23 skills that act as playbooks for specific kinds of work. It also includes a live dashboard that runs on your own computer and shows what the agents are actually doing in real time, plus a setup wizard called setup-forge that walks you through configuring API keys the first time you use it. There are three ways to install it: as a Claude Code plugin for a quick, read-only trial with no dashboard, through a one-line installer script for macOS, Linux, or Windows that sets up the full system, or by manually copying files into your project and your global Claude Code configuration folder. The project describes itself as having zero runtime dependencies, meaning it runs on plain Node.js scripts without needing npm install. The README emphasizes that the system tries to be honest about its own results: it keeps a ledger recording which agents actually ran on a given task, and states that nothing is reported as finished until it has been checked with real tests. claude-forge is released under the MIT license, which allows free use, modification, and distribution, including for commercial purposes.