What Happened

AI engineer Affaan Mustafa has open-sourced everything-claude-code (ECC), a production-grade enhancement layer for Anthropic's Claude Code CLI, according to a project breakdown published on Juejin. The project has accumulated 150,000+ GitHub stars and 23,700+ forks, with 170+ contributors, and is currently trending on GitHub.

ECC is not a standalone application. It installs on top of Claude Code CLI v2.1.0+ and augments it with 181 skills, 47 sub-agents, 34 rules, 8 lifecycle hook types, 79 legacy slash commands, and pre-configured MCP server integrations. Mustafa built it over 10 months of intensive Claude Code use, originating from a winning entry at an Anthropic hackathon.

The latest release is v1.10.0, dated April 2026. The project is MIT-licensed.

Why It Matters

For engineering teams already running Claude Code in production, ECC represents a community-aggregated configuration layer that would otherwise take months to build internally. The 150k + star count — if accurate — would place it among the fastest-growing developer tooling repositories of 2025-2026.

Three second-order effects are worth tracking:

  • Standardization pressure: With 170+ contributors codifying Claude Code best practices into a single repo , teams adopting ECC inherit community consensus on TypeScript, Python, Go, Rust, Swift, and eight other language rulesets simultaneously.
  • Multi -tool portability: ECC explicitly supports Cursor, Codex, OpenCode, and Gemini CLI in addition to Claude Code, positioning it as a vendor-neutral AI coding configuration standard rather than a Claude-specific tool.
  • Security baseline: The bundled AgentShield sub-agent (370 stars as a standalone project) introduces zero-width character injection detection and pre -tool-use security hooks — capabilities absent from Claude Code's default configuration.

The project description notes it is "splitting the developer community, " suggesting friction between teams that want opinionated AI tooling defaults versus those preferring minimal configuration.

The Technical Detail

Architecture: Six Modules

ECC layers six components over Claude Code's native . claude/ directory structure:

  • Skills (181): Categorized across backend frameworks (Django, Laravel, Spring Boot, NestJS), frontend patterns, CI/CD, multi-language specs (Python, TypeScript, Go, Java, Kotlin, Rust, C++, Swift), content creation, and media processing pipelines.
  • Agents (47): Specialized sub-agents including a Planner, Architect, Code Reviewer, Security Auditor, Build Error Resolver, E2E Test Generator, and language -specific experts for eight languages.
  • Rules (34): Enforced best-practice rules covering TypeScript, Python, Go, Swift , PHP, Perl, Java, Kotlin, C++, and Rust.
  • Hooks (8 lifecycle types): PreToolUse, PostToolUse, UserPromptSubmit, Stop, PreCompact, Notification, and SessionStart — each mapped to specific operational functions such as security checking, result logging, and memory persistence.
  • Commands (79): Legacy slash commands maintained for backward compatibility while skills migration continues .
  • MCP Configuration: Pre-wired integrations for GitHub, Supabase, Firecrawl, Vercel, Railway, Cloudflare, ClickHouse, Sequential Thinking, and a memory persistence MCP server.

Cross-Session Memory via Hook Composition

The most technically notable pattern in ECC is a three-hook composition using PreCompact + Stop + SessionStart to implement persistent memory across Claude Code sessions — a capability the source article notes is "almost entirely absent from Claude Code's official documentation."

  • PreCompact : Extracts and serializes key context before the context window is compressed.
  • Stop: On session end, saves the session summary and memory state to disk.
  • SessionStart: On new session initialization, loads the previously saved memory back into context.

This pattern effectively gives Claude Code a persistent working memory across disconnected sessions without requiring external database infrastructure.

Installation Profiles

ECC ships three installation profiles via a curl-pipe-bash installer:

  • minimal: Core rules + basic hooks only
  • standard: Adds skills and agents (documented as recommended)
  • full: All modules including experimental features
curl -fsSL https://raw.githubusercontent.com/affaan-m/everything-claude-code/main/scripts/install.sh | bash

Manual installation via git clone with selective copying of the .claude/ directory is also supported for teams requiring full configuration control.

Cost Optimization Features

The framework includes model ti ering strategies and MCP tool count thresholds — both aimed at reducing Claude API spend on high-volume agentic workflows. Specific pricing figures are not provided in available information.

What To Watch

  • Anthropic's official response: A third-party framework reaching 150k+ stars exposes gaps in Claude Code's native feature set. Watch for Anthropic to address cross-session memory, security hooks, or skill libraries in official Claude Code releases within the next 30 days.
  • Cursor and Gemini CLI adoption: ECC's explicit multi-tool support means adoption metrics will reveal whether developers are treating it as a Claude Code enhancer or a universal AI coding config layer. Cursor integration pull requests and issues will be a leading indicator.
  • AgentShield standalone trajectory: The bund led security scanner has 370 stars as a separate repo. Watch for it to be extracted into an independent security product or acquired as AI agent security becomes a compliance requirement.
  • v1.10.0 changelog details: The April 2026 release version suggests active development cadence. Review the changelog for breaking changes to hook APIs that could affect teams already running E CC in CI pipelines.
  • Enterprise fork activity: With 23,700+ forks, watch for enterprise-specific forks that strip experimental features and lock dependency versions — a signal that production adoption is scaling beyond individual developers.