ECC (Everything Claude Code) — What it is, and whether it’s worth it

If you have scrolled GitHub Trending lately, you have probably seen ECC (Everything Claude Code, canonical repo at affaan-m/ECC). Badges mention an Anthropic hackathon win and six-figure stars; headlines call it “the ultimate Claude Code config pack.” The real question is not hype—it is whether you should bolt the whole stack into your Cursor or Claude Code workflow, or treat it as a buffet. Below: what ECC actually ships, how to install without stepping on rakes, who wins, and where cloud gateways still matter.

What ECC is—and what it is not

The project brands itself a harness-native operator system. In practice that means an operations layer for coding agents: how context persists, how subagents run, how security gets checked—not merely longer system prompts. The upstream English README bundles skills, instincts, memory tuning, continuous learning, security scanning, and research-first workflows evolved from months of daily product work.

Distribution targets multiple harnesses—Claude Code, Codex, Cursor, OpenCode, Gemini, Zed, GitHub Copilot. v2.0.0-rc.1 adds Hermes operator flows, cross-harness architecture docs, and an in-tree ECC 2.0 alpha Rust control plane (ecc2/) with dashboard, sessions, and status. ECC is trying to fix agents that sprawl, get expensive, and grow unsafe—not to ship fifty more novelty slash commands.

Typical building blocks versus copying a single .cursorrules file:

• Agents for review, build repair, and architecture by language or scenario;
• Skills as hot-loadable workflows (tests, security, docs, ops);
• Hooks that summarize and persist context on SessionStart/Stop, gated by ECC_HOOK_PROFILE;
• Rules in per-language trees you install selectively;
• AgentShield via packages like ecc-agentshield and /security-scan-style commands.

Core repo is MIT; ECC Pro and the ECC Tools GitHub App monetize private-repo audits (~$19/seat/mo tier). Hobbyists can live on OSS; teams needing compliance trails may pay—two lanes, same upstream.

Install paths: plugin, scripts, npm—do not double-install

Many “is it worth it?” failures happen at install time. Official docs warn in bold: after /plugin install ecc@ecc, do not also run install.sh --profile full or npx ecc-install --profile full. You will duplicate skills and hooks—repeated commands, double-fired hooks, mysteriously bloated context.

Pick one lane:

• Claude Code marketplace: /plugin marketplace add https://github.com/affaan-m/ECC then /plugin install ecc@ecc;
• Selective clone + rules: plugins do not ship all rules/; copy only the languages you need;
• npm: ecc-universal for cross-tool and CI-oriented installs.

In Cursor, ECC adapts the same skills/rules across harnesses—still audit conflicts with project .cursor/rules and whether your org allows external hook scripts. Enterprise pattern: fork, pin a commit, publish an allowlisted skill catalog; do not track main weekly unless someone owns the merge.

Versus built-in Cursor / Claude Code features

Cursor in 2026 already ships strong agents: MCP, Rules, background agents, model routing. ECC’s value is versioning senior-engineer habits—git worktree parallelization, iterative subagent retrieval, checkpoint versus continuous eval loops, instinct import/export from real sessions.

Occasional autocomplete users will find ECC heavy. Teams pairing with agents four-plus hours daily, doing cross-repo refactors, or standardizing junior workflows get a checklist someone else already open-sourced. Stars measure attention, not that every skill fits your domain—treat the repo as a high-quality starter kit, not instant compliance.

Worth it? A decision table

Bottom line: worth using, rarely worth blind full install. Healthier rollout: minimal profile → one language ruleset → one memory hook → one week of observation. If latency, duplicate summaries, or bills spike, downgrade the hook profile before adding skills.

Security and maintenance

AgentShield, sandboxing, and CVE-oriented skills matter when agents run shell and read outside the repo. Still do your homework: read hook scripts; mirror internally for enterprise; never embed secrets in skill templates; read ECC Tools data scopes before installing the app. MIT means you can audit—not that uploads are automatically safe.

Upstream ships weekly; catalogs shift. Tracking HEAD is work; pinning a fork is calm. v2 operator skills (billing, workspace, social) may be noise for pure coders and gold for founder-operators—curate by role.

Split with OpenClaw, cloud Mac, and OpenHuman

ECC optimizes how agents behave on your machine. When you need always-on webhooks, IM channels, or headless browser queues, run OpenClaw Gateway on a Linux VPS—we covered that in GitHub Actions versus manual Docker for OpenClaw on Linux VPS. ECC does not replace TLS termination or persistent callbacks.

For desktop-long memory across mail and calendars, see OpenHuman’s Memory Tree approach—life-data sync versus ECC’s session operations hooks. xcodebuild, notarization, and TestFlight still require macOS; many teams keep build islands on cloud Mac and gateways on Linux. ECC owns “how we write and review”; hosts own “where we compile and listen.”

Summary

Yes if AI pairing is production for you and you will spend a day on minimal install, deduping rules, and tuning hooks—then maintain a pinned fork. No to full clone if you only need occasional completions or nobody owns hook policy—borrow a few skills instead.

Start at affaan-m/ECC, read the Shorthand Guide, then decide on full profile. Stars are social proof; your repo size, bill, and compliance bar decide whether ECC stays.