Agentkit.best (formerly Claudekit.cc) vs DIY and Free Kits

“Should I buy Agentkit.best or just build my own Claude kit?” is the most common question we get about the Claude kit category. This page compares the realistic options side by side: Agentkit.best, a do-it-yourself kit, the free Superpowers library, the unrelated open-source claudekit npm package, and Anthropic’s official plugin marketplace.

Facts below come from each project’s own site or README and the official Claude Code docs, checked on 24 October 2026. Seller figures (skill counts, prices) change between releases, so treat them as a snapshot. For a price-first view that also covers theclaudekit.com’s subscription kits, see Claude kit pricing.

Quick verdict

• Pick Agentkit.best if you want a large, opinionated plan-build-test-review workflow working today, across Claude Code and other agents, and you’d rather pay than spend days assembling one.
• Build your own if you want the leanest possible setup, full control, and no licence cost, and you have time to write and tune skills.
• Start with Superpowers if you want a free, well-known workflow (brainstorm, plan, TDD, subagent-driven development) without writing it yourself.
• Use carlrannaberg’s claudekit if you mainly want guardrail hooks and a handful of commands and subagents from an MIT-licensed npm package.
• Browse the official marketplace for single-purpose plugins you can mix into any of the above.

At a glance

	Agentkit.best	DIY kit	Superpowers	claudekit (carlrannaberg)	Official marketplace
Who makes it	Agentkit.best team	You	Jesse Vincent (obra)	Carl Rannaberg	Anthropic and partners
Price	Paid, one-time kit purchases	Free	Free	Free	Free to install
Licence	Commercial; site promises full source code ownership	Yours	MIT	MIT	Varies per plugin
What’s inside	108+ skills, 95+ slash commands, 45 subagents (seller figures)	Whatever you write	Workflow skills: brainstorming, plans, TDD, debugging, review	Hooks, slash commands, subagents	Individual plugins
Install	Kit download plus CLI setup	Files in .claude/ or ~/.claude/	/plugin install	npm install -g claudekit	/plugin install name@claude-plugins-official
Runtimes	Claude Code, Codex and others; Cursor in beta	Claude Code (skills also follow the open Agent Skills format)	Claude Code plus many other agents	Claude Code	Claude Code
Opinionated workflow	Strong	None until you add one	Strong	Light	Per plugin
Context footprint	Large library	As small as you keep it	Medium	Small to medium	Per plugin

What a Claude kit is made of

Every option here is a different packaging of the same building blocks, all built into Claude Code:

• CLAUDE.md for project facts and conventions, loaded every session. See the CLAUDE.md guide.
• Skills: folders with a SKILL.md that Claude loads when relevant or you run as /skill-name. See the skills guide.
• Subagents: Markdown files in .claude/agents/ with their own prompt, tools and model. See the subagents guide.
• Hooks: commands that run on lifecycle events such as PreToolUse and Stop. See the hooks guide.
• MCP servers for external tools. See best MCP servers.
• Plugins to bundle and share all of the above. See the plugins guide.

The real difference between kits is who writes these files, how many of them there are, and how strongly they push a particular workflow. If the terms are new, start with what a Claude kit is, and our skills vs subagents vs hooks explainer.

Agentkit.best

Agentkit.best (formerly Claudekit.cc) is the commercial kit; claudekit.cc now redirects to agentkit.best. Our Agentkit.best review covers it in more depth.

What the site lists (24 October 2026):

• 108+ skills across engineering and marketing.
• 95+ slash commands covering the delivery lifecycle, such as /ak:brainstorm, /ak:plan, /ak:cook, /ak:code-review and /ak:ship.
• 45 specialized subagents.
• Runtimes: Claude Code, Codex, Antigravity, Pi and Oh My Pi listed as production; Cursor and DeepSeek Harness as beta; OpenCode, Grok and GitHub Copilot as preview.

Pricing as listed on agentkit.best on 24 October 2026: Agentkit.best Engineer $99, Agentkit.best Marketing $99, or both as a $149 bundle, all one-time purchases. An optional desktop app is sold separately as a yearly licence. The site advertises a 14-day refund guarantee. Prices change, so check the site before you buy.

See current Agentkit.best plans

Disclosure: Claudekit.ai may earn a commission if you buy through this link. We’re independent and not affiliated with Anthropic or Agentkit.best’s makers.

Strengths

• Breadth: one install gives you planning, implementation, testing, review, docs and shipping workflows with named commands.
• The commands chain skills and subagents for you, which is the hardest part to design well yourself.
• Works across several agent runtimes, useful if your team mixes Claude Code and Codex.

Trade-offs

• It costs money while the underlying features are free.
• A large library means more skill descriptions sitting in context each session; Claude Code loads the name and description of every auto-invocable skill up front.
• Strong opinions: if your process differs, expect to override or disable parts of it.

DIY Claude kit

Building your own means writing the files above for your project and committing them, or packaging them as a private plugin for your team.

A minimal, effective DIY kit often looks like this:

CLAUDE.md                      # stack, commands, conventions
.claude/skills/release/SKILL.md
.claude/skills/review/SKILL.md
.claude/agents/test-runner.md
.claude/settings.json          # permissions + a formatter hook
.mcp.json                      # 1-3 MCP servers you actually use

Strengths

• Zero licence cost and nothing you don’t need, so the context footprint stays small.
• Every instruction reflects your codebase, not a generic process.
• Easy to review: it’s plain Markdown and JSON in your repo.

Trade-offs

• Time. Writing, testing and tuning good skill descriptions, subagent prompts and hooks takes iteration.
• You have to design the workflow yourself; that’s where most of the value of a mature kit sits.
• Maintenance is on you when Claude Code changes a field or event.

Anthropic’s skill-creator skill speeds up the writing part.

Superpowers

Superpowers by obra is an MIT-licensed skills library and methodology. Its README describes a flow of brainstorming before code, writing plans broken into small tasks, test-driven development, subagent-driven development with review checkpoints, and finishing a branch cleanly.

/plugin install superpowers@claude-plugins-official

Strengths: free, popular, focused on engineering discipline, and documented for many coding agents besides Claude Code.

Trade-offs: it’s a workflow library, not a full kit. You still bring your own project instructions, MCP servers and domain skills, and its process (TDD-first, plan-first) may feel heavy for small edits.

claudekit by carlrannaberg (different project)

Search results for “claudekit” also surface carlrannaberg/claudekit, an open-source npm package that is not related to Agentkit.best (formerly Claudekit.cc) or to this site. Its README describes “smart guardrails and workflow automation for Claude Code”:

npm install -g claudekit

It ships hooks (file protection, typecheck/lint/test on change, session context), slash commands such as /code-review, /spec:create, /git:commit and /checkpoint:restore, and expert subagents (TypeScript, React, testing, database and others). MIT licensed.

Strengths: free, guardrail-focused, light enough to sit alongside your own skills.

Trade-offs: Claude Code only, and a narrower scope than a full lifecycle kit.

Anthropic’s official plugin marketplace

Claude Code adds Anthropic’s official marketplace, claude-plugins-official, on your first interactive session. Browse it with /plugin, then install individual plugins:

/plugin install commit-commands@claude-plugins-official

Strengths: pick exactly the pieces you need, a more trusted source, and easy updates.

Trade-offs: it’s a catalog, not a workflow. Assembling plugins into a coherent process is still your job. Any plugin can run code with your permissions, so review hooks and MCP servers before enabling them. Details in the plugins guide.

The real cost: money vs time

We won’t invent a “hours saved” number, but the trade is easy to estimate for your own team:

1. List the workflows you want (plan, implement, test, review, release, docs).
2. Estimate how long it takes you to write and tune one good skill or subagent, then multiply.
3. Compare that with the kit price and the time you’ll spend learning and trimming the kit.

If the list is short, DIY or Superpowers usually wins. If you want most of the lifecycle covered across several runtimes, a paid kit becomes competitive quickly. Whichever you choose, watch your usage and context; our Claude Code pricing guide explains plan limits.

Choose Agentkit.best if

• You want a complete, named-command workflow on day one.
• Your team uses more than one agent runtime.
• You’d rather customize an existing system than design one.

Choose a DIY kit if

• You care most about a small context footprint and full control.
• Your workflow is unusual or tightly tied to your stack.
• You enjoy writing skills, or have someone on the team who does.

Choose Superpowers or claudekit if

• You want structure for free: Superpowers for a disciplined dev methodology, carlrannaberg’s claudekit for guardrail hooks and expert subagents.

Mixing options safely

Combining is fine if you avoid duplicate workflows. A sensible pattern: one main workflow source (Agentkit.best or Superpowers or your own), plus small project-specific skills, plus individual plugins from the official marketplace. Disable what you don’t use, since every enabled auto-invocable skill costs a little context on every turn.

Related

• Claude kit hub
• Agentkit.best review
• Agentkit.best: what you get and who should skip it
• Best Claude Code skills
• Claude Code vs Codex