Getting Started With ClaudeKit: Install a Kit in 5 Minutes

ClaudeKit installs in one of two ways — the ck CLI or the native Claude Code plugin system — and the whole process takes under five minutes. Five kits cover 101 commands and 82,197 measured tokens of context load. Pick the kit that matches the work you do this week, run ck auth <key> && ck install <kit>, and your first slash command is ready before your coffee cools.

What exactly is ClaudeKit and who is it for?

ClaudeKit is a set of five domain-specific command packs for Claude Code — Anthropic's agentic terminal. Each kit ships slash commands (workflows you invoke), skills (auto-loading knowledge files), and read-only specialist agents (reviewers, auditors, researchers). Nothing orchestrates anything without you; every command ends with a deliverable artifact — a diff, a report, a verified file — not a hand-off to another agent gate.

The five kits as of June 2026:

Token counts are measured, not estimated. Run ck tokens <kit> any time to recount against your installed version. We publish these numbers because context window cost is real — Claude Code demand is up 938% year-over-year as of June 2026, and every token in your context competes with every token your task needs.

If you are not yet a Claude Code user, install it first and spend a session in a real repo before buying anything. The kits extend a tool you already trust; they don't replace judgment.

How do you install ClaudeKit?

There are two install paths. Use whichever fits your workflow.

Path A: the ck CLI (recommended)

The claudekits npm package ships two binaries — claudekit and the short alias ck. You need Node.js 18 or newer:

node -v          # confirm 18+
npm i -g claudekits

Once installed, authenticate with your license key. The key arrives by email a few minutes after checkout and looks like ks_live_...:

ck auth ks_live_xxxxxxxxxxxxxxxx

This validates the key, registers one of your three allowed devices, and caches a token at ~/.claudekit/config.json. Run ck whoami to confirm which kits your license unlocks.

Then install the kit:

ck install seokit          # installs to current project: ./.claude/
ck install seokit --local  # same, explicit
ck install seokit          # add no flag for project default

To install globally — available in every Claude Code session regardless of directory:

ck install seokit -g   # writes to ~/.claude/

Project installs override global ones when both exist, so you can pin different kit versions per repo. The CLI prints the kit's full token ledger as files land. That ledger is the exact byte count that will load into your context window on each session.

Path B: the native plugin marketplace

If you prefer zero CLI setup, use Claude Code's built-in plugin system:

/plugin marketplace add Madni-Aghadi/claudekit-seo

Replace seo with eng, mkt, video, or ecom depending on which kit you bought. The plugin system handles auth via your Anthropic account; your ClaudeKit license key links in the dashboard at checkout.

Both paths install the same files. The CLI gives you ck doctor, ck list, ck tokens, and ck update maintenance commands. The plugin path gives you one-command installs directly inside Claude Code.

What happens after install — your first command?

Open Claude Code in the same directory where you installed, then type the kit's namespace:

/seo

Typing a bare namespace shows every command the kit ships with one-line descriptions. Pick one and run it. For SEOKit, a practical first command:

/seo audit https://yoursite.com

This runs a structured technical and content audit, writes findings to a markdown file in your project, and returns a prioritized fix list. The output is the deliverable — no follow-up agent, no approval step.

For EngineerKit, the daily driver is the "Daily-8" sequence:

1. /eng catchup — summarize what changed since last session
2. /eng plan — break a task into steps with risk flags
3. /eng tdd — generate failing tests first
4. /eng debug — root-cause-first diagnosis (the flagship command)
5. /eng verify — confirm the fix holds
6. /eng review — self-review before pushing
7. /eng commit — conventional commit with context
8. /eng handoff — write a handoff note for async teammates

The /eng debug command is our most-used command across all kits. It forces root-cause reasoning before proposing a fix — no pattern-matching to the nearest Stack Overflow answer.

Which kit should you start with?

Match the kit to the work you do most often this week. Here is a quick routing guide:

A few notes on the less-obvious flagships:

/mkt voice reads a sample of your actual writing and produces a voice file — a structured description of your tone, sentence patterns, vocabulary, and cadences. Every subsequent MarketingKit command loads that file automatically. /mkt humanize then strips 14 documented AI tells from any draft. Together they close the gap between "AI-written" and "sounds like me."

/seo quick-wins pulls your Search Console data, finds positions 8-20 with decent impressions, and finds low-CTR pages where a title rewrite would move the needle immediately. AI Overviews now appear on 48% of Google queries (March 2026, up from 34.5% in December 2025), so positions 8-20 that get AIO coverage convert better than their traditional click-through rate suggests — this command finds those pages.

/ecom no-sales runs your store against AOV-band benchmarks and flags which lever is most likely broken: traffic, conversion rate, AOV, or repeat rate. It returns a triage report, not a guess.

/video clone takes a reference video, reconstructs its style in a Remotion template, and verifies the match. The VideoKit skill layer understands Remotion's component model, so the output actually runs.

What are the pricing options?

ClaudeKit uses three tiers, all with a 14-day refund window and a 3-device limit per license:

The lifetime option is one kit as it ships at purchase — no update stream included. If you want updates, the monthly or annual plans are the right call. The Pro tier is useful if your work rotates: a developer who occasionally runs campaigns can hold EngineerKit and MarketingKit without paying for SEOKit. See the full pricing page for the current comparison.

What is the token cost and does it matter?

Yes, it matters. Skills and agents load into your context window on every session where they are installed. EngineerKit's 20,413 tokens is about $0.06 per session at Sonnet 4.6 rates — low enough to ignore individually, but worth knowing if you are running long agentic loops.

We have measured the full token load for every kit at install time (see the table above). ck tokens <kit> recounts on demand against your local files. No other kit publisher we are aware of publishes per-install token counts. The reason is simple: hidden context load compounds fast, and we would rather you know the cost upfront.

For context: Anthropic shipped Opus 4.8 in May 2026 at $5/$25 per million tokens with a 3x cheaper fast mode. At those rates, running all five kits in the same session (82,197 tokens) costs about $0.41 in context load per session at Opus 4.8 input pricing — real money if you run 50 sessions a day, trivial if you run five.

What if something breaks?

Most issues fall into three categories:

1. ck command not found after install — your npm global bin directory is not on PATH. Run npm bin -g to see where it went, then add that path to your shell config.
2. Commands do not appear in Claude Code — you installed in a different directory than the one your Claude Code session is open in. Either reinstall where you work or use ck install <kit> -g for the global location.
3. Key auth fails — the key has already been used on three devices. Use ck whoami to see registered devices, then contact support to deregister an old machine.

For anything else, ck doctor runs a self-diagnostic and prints what it finds. The docs cover the full CLI reference. The 14-day refund policy means a kit that genuinely does not fit your workflow costs nothing to return — email from your purchase address.

FAQ

Do I need to install all five kits?

No. Each kit installs and runs independently. Most users start with one, run it for a week, and add more if they want coverage in another domain. The Pro plan lets you hold any three kits and swap one per billing cycle, which is a practical middle ground if your work spans two or three domains.

Can I use ClaudeKit on a work machine and a personal machine?

Yes. Each license covers three devices. ck auth <key> on each machine registers that device. If you need to move to a fourth machine, contact support to deregister one of the existing three.

What is the difference between commands, skills, and agents?

Commands are slash workflows — you invoke them explicitly and they run a multi-step task. Skills are knowledge files that auto-load into Claude's context when relevant, giving it domain-specific background without you typing it. Agents are read-only specialists — reviewers, auditors, researchers — that Claude can call during a command run. No agent in any kit can trigger another agent or block a command from completing. Every command ends with an artifact, not a gate. See the full explainer for the architecture breakdown.

Is ClaudeKit compatible with Claude Code teams plans?

Yes. If your team runs Claude Code under an Anthropic teams subscription, ClaudeKit installs the same way — either via ck install or /plugin marketplace add. Each team member needs their own ClaudeKit license. All-Access annual at $399/year per seat is the most common option for teams.

What happens when kits are updated?

Monthly and annual subscribers get all updates automatically. Run ck update <kit> to pull the latest version; the CLI prints a changelog summary and the new token ledger. Lifetime license holders receive the kit as it was at purchase — no update stream.

How is ClaudeKit different from free prompt packs?

Free prompt packs are static text files. ClaudeKit ships structured slash commands with multi-step logic, skills that auto-load domain knowledge, and agents that run read-only specialist tasks. The token cost is measured and published. Commands produce artifacts (diffs, reports, verified files) not conversational replies. The comparison post goes deeper on the practical differences.

If you write and ship code most days, EngineerKit is the fastest way to see what ClaudeKit does — install it, run /eng debug on a real bug, and watch it work root-cause-first rather than guess-and-check. If content and distribution is where you spend your time, MarketingKit with /mkt voice and /mkt humanize will change how fast you produce work that sounds like you. Either way, the install takes five minutes and the 14-day refund window means the only real cost is the session you spend trying it.