Everything Claude Code (ECC) has become the dominant open-source agent harness for AI coding assistants, surging past 262,000 GitHub stars with 837 new stars gained in a single day — a pace that puts it among the platform's fastest-growing repositories.
Created by Affaan Mustafa over 10 months of daily use, ECC began as a personal configuration pack for Claude Code and has since evolved into what its maintainer calls an "agent harness operating system" — a cross-platform framework that provides skills, instincts, memory management, security guardrails, and workflow tools for AI coding agents.
ECC 2.0, released this year, marked the project's graduation from a Claude Code-centric toolkit to a multi-harness OS. It now supports Claude Code as its first-class citizen while also providing adapters for OpenAI's Codex, Cursor, and other coding agents. The system is entirely file-based — markdown and YAML — with no build step required.
The project's architecture centers on three concepts: skills (domain expertise Claude draws on automatically), hooks (triggered actions at defined points in the workflow), and instincts (behavioral rules that shape how agents operate). Together, they bolt a full engineering process — planning, execution, review, memory — onto whichever coding agent the developer chooses.
The repo's explosive growth reflects a broader trend: as AI coding agents move from novelty to daily-driver status, developers are investing heavily in configuring and optimizing them. ECC provides the community-maintained baseline that turns a raw coding assistant into a structured engineering partner.
With 262K+ stars, ECC now rivals the star counts of major language runtimes and framework repos, signaling that the AI agent tooling ecosystem has reached a new order of magnitude.




