Skip to main content
PromptQuorumBuilt for humans. Structured for AI.
Home/Power Local LLM
Local AI Agents & Tool Use

ECC Review: Agent Toolkit for Claude Code and Codex

·8 min readBy Hans Kuepper · Founder of PromptQuorum · Discovery engine for open-weight & open-source AI

ECC is an MIT-licensed toolkit of agents, skills, hooks, rules and memory that you install into a coding agent such as Claude Code or Codex, plus a separate hosted GitHub App. It adds a structured engineering workflow to an agent you already have; it does not supply a model.

Key Takeaways

  • The toolkit costs nothing; only the optional GitHub App has paid plans
  • ECC runs no model of its own; which model answers depends on the agent it is installed into
  • Setup is one terminal command that needs Node.js 18 or newer, not a double-click installer
  • The GitHub App sends commit data to Anthropic's Claude API for analysis, according to ECC's own privacy policy
  • Usage figures and customer logos on the website are self-reported and were not independently verified

📍 In One Sentence

ECC is a free, MIT-licensed add-on that gives your coding agent a ready-made set of agents, skills, hooks and memory.

💬 In Plain Terms

ECC does not write code itself. You install it into the coding agent you already use, and it adds reusable workflows for planning, reviewing and remembering what happened. A separate paid service, the GitHub App, reads a repository's history and proposes guidance.

Read in:
🇺🇸en🇩🇪de🇫🇷fr🇯🇵ja🇨🇳zh🇪🇸es🇧🇷pt🇸🇦ar🇰🇷ko

ECC (ecc.tools, source at github.com/affaan-m/ECC) is an open-source toolkit of agents, skills, hooks, rules and memory that you install into a coding agent such as Claude Code or Codex. A separate, paid GitHub App turns a repository's history into reviewable guidance. It belongs in a local-AI directory because the toolkit runs inside your own agent workflow, but it runs no model itself. This review covers what ECC installs, how setup works, what the GitHub App sends to third parties, what it costs and who it fits.

ECC Review: Agent Toolkit for Claude Code and Codex

What Is ECC?

ECC is an open-source project that packages a software-engineering workflow as installable parts for coding agents: plan, test, implement, review, verify, remember and improve. The repository describes itself as an "agent harness performance optimization system," and the website at ecc.tools adds a hosted GitHub App on top.

  • Author: Affaan Mustafa, credited on the website footer; the README describes a single maintainer releasing weekly
  • Repository: github.com/affaan-m/ECC, MIT licence, verified via the GitHub API
  • Packages: ecc-universal (the toolkit installer) and ecc-agentshield (the security scanner), both MIT on npm

What Does ECC Install?

According to the README for v2.2.3, ECC provides 71 agents, 311 skills and 97 legacy command shims, plus hooks, rules, memory and AgentShield. These are the README's own counts, not ones PromptQuorum checked.

  • Agents and skills — reusable roles and procedures for planning, code review, build repair, security review, test-driven development, research and documentation
  • Hooks and memory — runtime enforcement, session summaries, continuous learning that turns repeated work into reusable skills, and context controls
  • Rules — always-loaded coding standards you pick per language or project; Claude Code plugins cannot distribute rules, so they are copied by hand
  • AgentShield — a scanner for prompts, hooks, MCP configuration, permissions, secrets and agent files; the website calls a scan a review aid, not a guarantee of safety
  • Plan Canvas and Control Pane — a browser view for annotating a plan and a local view of agent sessions and worktrees, shown on the website with demonstration data

Which Coding Agents Does ECC Support?

According to its README, ECC works best with Claude Code, has native plugin paths for Codex and Qoder, and offers adapters with fewer capabilities for other tools. Choose one install method per agent; the README warns that stacking methods can duplicate skills and hooks.

Claude Code

Support level:
Native plugin ecc@ecc; the README's main path

Codex

Support level:
Native plugin through the guided setup

Kimi Code

Support level:
Managed project files; hooks, model settings and sign-in are not configured

Qoder

Support level:
Native plugin path, per the README

Cursor, OpenCode, Gemini, Zed, GitHub Copilot, Antigravity, Qwen

Support level:
Capability-limited adapters installed with ecc install --target

The website lists Claude Code, Codex, Cursor, OpenCode, Kimi Code and GitHub as platforms; the README names the longer adapter list above.

The GitHub App and Your Data

The GitHub App is ECC's hosted repository service, and its privacy policy says commit data is sent to Anthropic's Claude API for analysis. The policy, dated March 2026, covers the app; it does not describe what the coding agent you pair ECC with sends to its own model provider.

  • Purpose: the app reads a repository's git history, generates SKILL.md files and repository "instincts," and opens pull requests; the terms say it runs when you trigger it with an issue comment
  • What it accesses: repository metadata, commit messages, author names, dates, file change summaries, limited code samples and GitHub installation data
  • Third parties: Anthropic's Claude API for pattern analysis, the GitHub API, and Cloudflare for hosting
  • Retention: analysis results are cached for up to 7 days; code samples are processed in memory and not stored permanently; usage counts are kept monthly for rate limiting

ECC Pricing

The ECC toolkit is free under the MIT licence; only the GitHub App has paid plans. Prices below come from ecc.tools/pricing and are listed in US dollars only.

  • Yearly billing saves about 17 percent on Pro, according to the pricing page
  • Pro offers optional metered overage for heavier use; the free plan needs no credit card

Free

Price:
$0
What it covers:
Public repositories only, 10 analyses per month, 200 commits per run

Pro

Price:
$19 per seat per month
What it covers:
Private repositories, 50 analyses per active seat pooled on the installation, 1,000 commits per run, AgentShield-backed scanning

Enterprise

Price:
Custom
What it covers:
Unlimited analyses, 5,000 commits per run, SSO-ready governance, policy packs, onboarding

Check the checkout page for current figures; this review did not buy a plan.

What We Could Not Verify

This review rests on the website, the GitHub repository, the npm registry and ECC's published policies. PromptQuorum did not install or benchmark ECC.

  • Whether ECC works with a locally hosted model: ECC does not document this, and the answer depends on the agent you install it into
  • The 71 agents, 311 skills and 97 command shims: these are README counts, not counted here
  • Traffic and adoption figures, and the customer logos shown on the website: self-reported by the project
  • Where the Unified Memory feature stores its data: the website does not say

Who Should Use ECC?

ECC fits developers who already work in a coding agent and want a prepared workflow instead of writing their own skills and hooks.

Use ECC if you

  • Work daily in Claude Code or Codex and want planning, review and security skills pre-built
  • Switch between several coding agents and want one shared set of skills and memory
  • Are comfortable reading hook and plugin code before running it

Look elsewhere if you

  • Want a standalone coding assistant that brings its own model; ECC needs an agent underneath
  • Need every component to run offline with a local model, which ECC does not document
  • Cannot send repository history to a third-party AI provider, which rules out the GitHub App

Competitors and Alternatives

ECC is an add-on layer, so the nearest same-category reviews are coding agents it can sit on, or that offer a comparable workflow of their own. None of them bundles ECC's mix of skills, hooks and memory in one package.

See the Local LLM Software Directory for the full catalog, including ECC's own entry.

Compare in directory

Frequently Asked Questions

Is ECC made by Anthropic?

No. The website credits Affaan Mustafa as author. Anthropic appears only as a third-party service that the GitHub App sends data to.

Do I need the GitHub App to use the toolkit?

No. The pricing page says the open-source tools are available independently of the GitHub App plans.

Does ECC work on Windows?

The README includes a PowerShell walkthrough for the guided setup, and the guided path does not require cloning the repository.

Can I install ECC twice?

You can use it in several agents at once, but choose one install method per agent. The README warns against combining the plugin with a full manual install.

How do I remove ECC?

The README has a Reset / Uninstall section. Uninstalling the GitHub App is done from your GitHub account.

Sources

← Back to Power Local LLM

MCP