ECC (Everything Claude Code)

Persistent agents, skills and memory for Claude Code users tired of re-explaining their workflow

Category
AI agents & assistants
Audience
Developers

Updated

ECC, short for "Everything Claude Code", is a free plugin layer that turns an AI coding assistant into a standing multi-agent development system with its own sub-agents, reusable skills, hooks and memory. It is aimed at developers who already work inside Claude Code (or OpenAI Codex, or Qoder) every day and are tired of restating their process, standards and project context in every new session. It was written by a single developer, Affaan Mustafa, and released under the MIT licence on 2026-01-18 according to its README.

What it does

ECC bolts onto a bare coding assistant and installs what its author calls an operating system on top of it. Instead of one session that forgets everything on exit, you get a cast of specialized workers and a library of procedures the assistant can reach for.

By the README's own count, ECC ships:

  • 71 specialized agents
  • 311 reusable skills
  • 97 legacy "command shim" entry points, kept for teams moving to the skills-first model

Primary support is for Anthropic's Claude Code, with native support for OpenAI Codex and Qoder, plus adapters for Cursor, OpenCode, GitHub Copilot and Kimi Code. The same written process travels with you if you switch assistants.

How it works

ECC imposes one fixed workflow on whatever you ask for: plan, test, implement, review, verify, remember, improve. Each stage can hand off to agents built for it, and the skills are the reusable pieces those agents call rather than prompts you retype. Hooks and a memory store are what carry decisions from one session into the next, which is the whole point of the project: vanilla Claude Code starts cold every time.

The older command entry points still exist as shims so an existing team setup does not break while it migrates to skills.

Getting started

ECC is a layer, not a model, so the first requirement is an assistant it can attach to: Claude Code, Codex or Qoder natively, or one of the adapted tools. Whatever account, subscription or API key that assistant needs is still on you, and the materials here do not spell out those costs. The project itself does not list operating system, chip, memory or VRAM requirements, so treat "it runs where your assistant runs" as the honest answer rather than a promise.

ECC itself is free and MIT-licensed. There is one paid tier: an optional "ECC Pro" GitHub App at 19 dollars per active seat per month for private-repository coverage, with custom Enterprise pricing for rollout, training and security-policy support, as listed on ecc.tools.

Installation is a plain command, not a download from Releases. The README gives two routes:

  • npx ecc-universal install --guided for the universal installer
  • /plugin install ecc@ecc inside Claude Code

The repository is at github.com/affaan-m/ECC.

When to use it / when not

The main catch is time. Independent coverage from dev.co, cited by the project and repeated in the video, puts a proper setup at two to four weeks. That is not a lunch-break install: you are adopting a process, tuning agents to your codebase and deciding which of 311 skills your team actually uses.

Use it instead of plain Claude Code when you keep pasting the same conventions, test rules and review checklist into every session, and when more than one person needs that process to be identical. Do not use it if you dip into an assistant occasionally for one-off scripts, or if you want results today: a hand-written project instruction file will cost you an afternoon instead of a month.

One more caution about the headline figure. The 281,396 stars and 41,839 forks are the README's own claim as of 2026-10-16, for a repository it says was created on 2026-01-18. Popularity numbers are not a measure of how well the harness works on your code, and nobody in these materials has benchmarked that.

Alternatives

The comparison the reader already knows is vanilla Claude Code: Anthropic's own CLI, used session by session with no persistent scaffolding. ECC does not replace it, it sits on top of it. The cheaper middle ground is writing your own project instructions and a handful of custom commands, which costs almost nothing and gives you a fraction of the structure. ECC is the maximal version of that same idea, with the setup bill that implies.

If you drive a coding assistant for hours a day on a codebase that outlives any single chat, ECC is worth a serious look: it is MIT-licensed, installable in one command, and portable across Claude Code, Codex and Qoder, so the process you write down is not locked to one vendor. If your use is casual, or you cannot spare the weeks dev.co says a real rollout takes, skim the agent and skill lists for ideas and keep using the assistant as it comes.

More in AI agents & assistants

All of AI agents & assistants →