affaan-m/everything-claude-code

⭐ 177,399  ·  JavaScript  ·  GitHub Repo

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

ai-agents anthropic claude claude-code developer-tools llm mcp productivity

1-Sentence Summary

A production-tested performance optimization ecosystem for AI agent harnesses, evolving skills, memory, and security across platforms.

🔥 Key Capabilities & USP

• Memory Persistence & Continuous Learning — Automatically saves and loads context across sessions, extracting patterns from past work into reusable skills. Solves the painful problem of agents forgetting everything between sessions, forcing you to re-explain your project state repeatedly.
• Token Optimization Engine — Slims system prompts, intelligently selects models, and runs background processes to reduce token consumption. Directly addresses the high cost of running AI agents at scale by cutting unnecessary API usage.
• AgentShield Security Scanning — Built-in attack vector detection, sandboxing, sanitization, and CVE monitoring integrated into the agent workflow. Solves the critical security gap most agent harnesses have—running untrusted code or prompts without guardrails.
• Cross-Harness Compatibility — Works seamlessly across Claude Code, Codex, Cursor, OpenCode, and Gemini with behavior-tightened parity. Eliminates vendor lock-in and lets you switch agent platforms without rebuilding your optimization setup.
• Selective Install Architecture — Manifest-driven install pipeline with state tracking for targeted, incremental component installation. Solves the "bloated setup" problem by letting you install only what you need, when you need it.

USP: Unlike simple configuration templates or single-harness tools, Everything Claude Code is a complete, production-tested ecosystem evolved from 10+ months of intensive real-world use, with 48 agents, 182 skills, and 997+ internal tests—all working across multiple AI agent platforms.

Technical Architecture

Layer	Technology	Purpose
Core Runtime	JavaScript (Node.js), TypeScript	Primary agent harness integration and hook system
Component Library	Python, Go, Java, Perl, Rust, Kotlin, C++, PHP	Multi-language skill execution and agent support
State Management	SQLite	Session persistence, install manifests, state tracking
Desktop Dashboard	Tkinter (Python)	GUI for monitoring and managing ECC components
Control Plane (Alpha)	Rust	ECC 2.0 next-gen performance and daemon management
Agent Integration	MCP (Model Context Protocol), custom hooks	Claude Code, Codex, Cursor, OpenCode, Gemini harnesses
Testing	997+ internal tests	Full suite green for production reliability

Quick Start Guide

# Clone the repository
git clone https://github.com/affaan-m/everything-claude-code.git
cd everything-claude-code

# Launch the desktop dashboard (Tkinter-based)
python ecc_dashboard.py
# OR
npm run dashboard

# For ECC 2.0 alpha (Rust control-plane)
# Available commands:
# dashboard, start, sessions, status, stop, resume, daemon

# Configure hook runtime behavior
export ECC_HOOK_PROFILE=standard  # Options: minimal | standard | strict
export ECC_DISABLED_HOOKS=hook_name1,hook_name2

# Set up for Codex CLI compatibility
/codex-setup  # Generates codex.md

# Available harness commands
/harness-audit    # Audit current harness configuration
/loop-start       # Start continuous improvement loop
/loop-status      # Check loop status
/quality-gate     # Run quality checks
/model-route      # Route to optimal model

Pros, Cons & Use Cases

Pros

• Production-tested over 10+ months of intensive real-world use with 997+ passing tests
• Cross-harness compatibility eliminates vendor lock-in for agent platforms
• Comprehensive security with AgentShield attack detection, sandboxing, and CVE monitoring
• Active community with 177K+ stars and 170+ contributors indicating strong adoption
• Multi-language support across TypeScript, Python, Go, Java, Rust, and more

Cons

• ECC 2.0 alpha (Rust control-plane) is still in development—not production-ready
• Complex system with 48 agents, 182 skills, and 68 legacy command shims requires significant learning investment
• Heavy dependency on Tkinter for the dashboard, which may not suit all environments
• Documentation overhead to understand the selective install architecture and manifest system

Who should NOT use this?

• Casual users who only need a simple configuration file for a single agent session
• Teams on a tight deadline who cannot invest time in learning a complex ecosystem
• Developers using only one agent harness with no plans to switch or integrate multiple platforms
• Projects with minimal security requirements where AgentShield's comprehensive scanning is overkill

Ideal Use Cases

• Enterprise AI development teams using multiple agent harnesses (Claude Code, Codex, Cursor) who need consistent behavior and security across all platforms
• Long-running agent workflows that require memory persistence and continuous learning across sessions
• Security-conscious organizations deploying AI agents for code generation and execution who need built-in attack vector detection
• Power users who want to optimize token consumption and reduce API costs at scale
• Research teams exploring agent behavior across different harnesses and needing a unified optimization framework

Community & Activity

Everything Claude Code has explosive momentum with 177,399 stars and 170+ contributors—a testament to its real-world value. The project is actively maintained with the latest update on May 10, 2026, showing ongoing development. The community has already contributed 997+ internal tests with full suite green, indicating strong quality standards. With ECC 2.0 alpha in development (Rust control-plane), the project is clearly evolving toward even better performance. This isn't just a side project—it's a rapidly growing ecosystem with serious adoption and a vibrant contributor base pushing it forward.