everything-claude-code·Beginner·Last tested: 2026-04·~5 min read

Everything Claude Code

Everything Claude Code is an agent harness performance optimization system built for Claude Code, Codex, Cursor, and other AI development tools. This system provides skills, memory persistence, security scanning, and continuous learning capabilities evolved from 10+ months of production use.

Key Features

Agent performance optimization — Token optimization, memory persistence, and continuous learning loops
Security scanning — AgentShield component with vulnerability detection and sanitization
Cross-platform support — Works with Claude Code, Cursor, Codex CLI/app, and OpenCode
Skills library — 50+ specialized skills for TypeScript, Python, Go, Java, Kotlin, and 7 other languages
Memory system — Hooks that automatically save/load context across sessions
Verification loops — Checkpoint and continuous evaluation with grader metrics
Subagent orchestration — Git worktree parallelization and cascade methods

Installation

Install the universal package:

npm install -g ecc-universal

For security features:

npm install -g ecc-agentshield

Clone and run selective install:

git clone https://github.com/affaan-m/everything-claude-code
cd everything-claude-code
node install-plan.js
node install-apply.js

Info

Use ECC_HOOK_PROFILE=minimal|standard|strict and ECC_DISABLED_HOOKS=... to control runtime behavior without editing hook files.

Basic Usage

Initialize in your Claude Code workspace:

# Setup for Claude Code
ecc init --harness claude-code

# For Codex CLI
ecc codex-setup

# Enable specific skills
ecc skill enable typescript-reviewer
ecc skill enable memory-persistence

Basic harness commands:

// In your AI chat
/harness-audit    // Performance scoring
/loop-start       // Begin continuous eval
/quality-gate     // Checkpoint validation
/model-route      // Switch model strategies

Tip

Start with the Shorthand Guide on the project's social media for setup fundamentals, then reference the Longform Guide for advanced token optimization.

Notable Details

License: MIT
Language: JavaScript with TypeScript, Python, Go, Java support
Community: 135K+ stars, 6K+ forks, 30+ contributors
Languages supported: 7 (English, Portuguese, Chinese, Japanese, Korean, Turkish)
Origin: Anthropic Hackathon Winner

The project includes 997 internal tests and maintains compatibility across major AI development harnesses through a manifest-driven install system.