Everything Claude Code

Everything Claude Code is a comprehensive performance optimization system for AI agent harnesses, originally built at an Anthropic hackathon and refined over 10+ months of intensive production use. With 75,900 GitHub stars and 9,500 forks, it's one of the most widely adopted Claude Code setups in existence. The system bundles 12+ specialized subagents (planner, architect, security reviewer, code reviewer, debugger, and more), 65+ reusable skills spanning backend patterns, frontend frameworks, testing, security, and domain knowledge, plus 40+ slash commands for common operations. Memory persistence lets agents learn from previous sessions, and verification loops catch bugs before they reach production. Everything is tested with 997 internal tests across 5 language configurations: TypeScript, Python, Go, Java, and Perl. The core insight behind the project is that most developers run Claude Code with default settings and leave 80% of its capability unused. Token optimization, smart parallelization, security-first verification, and research-first development are all disabled by default. Everything Claude Code turns all of it on, in a carefully tuned configuration that doesn't generate noise or slow down your workflow. Cross-platform support extends to Cursor, OpenCode, Codex CLI, and standard terminal environments on Windows, macOS, and Linux.