Best Claude Code Kits & Plugins 2026: Ranked (AgentKit #1)

Prices and GitHub star counts change - verify live before you buy or clone anything below.

⚠️ Which AgentKit? (not OpenAI AgentKit)

Before anything else, a name collision that sends people to the wrong product. When you search "AgentKit," Google mostly shows OpenAI AgentKit - the Agent Builder, ChatKit, and Connector Registry suite OpenAI announced on October 6, 2025 (OpenAI, introducing AgentKit, 2025-10-06). That is a platform for building your own agents, and it is not what this ranking is about.

The #1 pick here is a different product with the same name: AgentKit for Claude Code, at agentkit.best (accessed 2026-08-09), installed and run through the ak command-line tool. It is a curated bundle of skills, subagents, commands, and workflows that plug into Claude Code (and other AI coding CLIs). There are also unrelated name-twins on GitHub - agentkits.net, JimLiu/claude-agent-kit, and the older docs.claudekit.cc (AgentKit is the successor to ClaudeKit). If a page mentions "Agent Builder" or "ChatKit," you are looking at OpenAI's product, not this one.

What counts as a "Claude Code kit"? (kits vs plugins vs skills)

The rankings below mix three things that people lump together, so let me define them first - otherwise it is apples to oranges.

• Skill packs - free repositories full of Claude Code skills (self-contained instruction folders Claude loads on demand). You clone them into .claude/skills/ yourself. Example: anthropics/skills.
• Plugins - installable bundles from a marketplace (the official Anthropic marketplace or a community one). One command installs them, and they can carry skills, subagents, hooks, and slash commands together. Example: pr-review-toolkit.
• Kits - a curated, maintained collection of many skills + subagents + commands + workflows sold or distributed as one product, usually with an installer and updates. Example: AgentKit. A kit is the "batteries included" option; a skill pack is "assemble it yourself."

All three ultimately do the same thing - they extend what Claude Code can do without you writing every prompt from scratch. The difference is curation, maintenance, and how much you assemble by hand. That distinction is the whole point of this ranking: some people want a tidy free repo, others want a paid kit that is kept current for them. If the vocabulary here is new, start with what Claude Code is and the difference between skills, subagents, hooks, and MCP.

How we ranked them (scoring criteria)

Most "best of" lists rank by GitHub stars or gut feel. Stars measure popularity, not whether a kit is worth installing today. Here is the rubric I scored every entry against, weighted out of 100:

• Coverage & breadth (25) - how many real workflows it handles (frontend, backend, database, DevOps, review, docs, marketing) versus one narrow trick.
• Quality & maintenance (25) - are the skills well-written and actively updated, or abandoned?
• Setup speed (15) - minutes from "download" to "working in Claude Code."
• Docs & support (15) - is there real documentation, or just a README?
• Value (20) - what you get for the price, where free counts as automatically high on cost but is judged on what you have to assemble yourself.

Two honesty rules: a paid kit does not get bonus points for being paid, and a free repo does not get bonus points for being free. Value is graded on the whole picture - a free repo you spend a weekend wiring together is not "free" in time. Star counts and last-commit dates below are indicative and shift; treat them as a signal, not gospel.

What I deliberately left out: single-file gists, abandoned repos with no commits in the last six months, and "prompt collections" that are just markdown you paste manually - those are not kits or plugins in any useful sense. I also excluded OpenAI AgentKit and other agent-building platforms, because this ranking is about extending Claude Code, not building agents from scratch. Everything below is something I would actually install this week.

The best Claude Code kits & plugins for 2026 - ranked

Here is the shortlist at a glance, then the full write-ups. "Type" tells you whether it is a paid kit, a free skill pack, or a free plugin. Scores use the rubric above; ties were broken on maintenance and setup speed, because a slightly less broad kit that installs cleanly beats a sprawling one you fight for an hour. One note before the table: exactly one paid product made this list. That is not because paid kits are rare - it is because most of them are thin wrappers around free repos, and AgentKit was the one whose curation and maintenance clearly earned the price. If that changes on your live check, re-score it honestly.

AgentKit (agentkit.best) · best paid all-in-one

What it is: A commercial, curated kit for Claude Code (also works with Codex and Copilot CLIs) installed through the ak command. Platform totals are 108+ skills, 95+ commands, and 45 AI agents - 17 engineer agents and 28 marketing agents. It splits into an Engineer Kit (60+ skills, 30+ workflows across frontend, backend, database, DevOps, and code review) and a Marketing Kit (12+ MCP integrations, 3 workflows for lead gen, email/SEO, and CRM).

Best for: Developers who use Claude Code daily and would rather pay once than spend weekends stitching free repos together and keeping them current. In practice that means the person who has already tried cloning three GitHub repos, hit version drift, and decided their time is worth more than the friction.

The Engineer Kit is the one most readers here will care about: it covers the full build cycle - scaffolding frontend and backend, database work, DevOps, and code review - as maintained skills and workflows rather than one-off prompts. The Marketing Kit is a genuinely different product aimed at growth and content work (lead gen, email/SEO, CRM), so buy the Bundle only if you actually do both. Do not pay $149 for the marketing half if you are a pure engineer.

Price / license: Engineer Kit $99, Marketing Kit $99, Bundle $149. The site lists a money-back guarantee (no specific duration stated) and lifetime updates for the kits; the $99/$149 prices are shown with no recurring fee listed (the separate Desktop App is the subscription piece, at $19/year early access). Source: agentkit.best, accessed 2026-08-09.

Pros: Widest coverage of anything here; the ak installer is a native binary (macOS/Linux/Windows) that needs no Node or Bun and sets up in under 5 minutes; skills are maintained and versioned; one license, no assembly.

Cons: It costs money when strong free options exist; you still need Claude Code (and a plan) first; and some of its skills overlap what you can get free from anthropics/skills or superpowers. It is not magic - it is curation and maintenance you are paying for.

Verdict (91/100): Earns #1 on coverage and setup speed, not on hype. If your time is worth more than $99, it pays for itself fast. Read our full AgentKit review and the AgentKit Engineer Kit deep-dive before you buy.

obra/superpowers · best free skills framework

What it is: An open-source framework of Claude Code skills and workflows that has become the reference free setup for power users. Best for: Developers comfortable cloning a repo and wiring skills into .claude/ themselves. Price: Free, open source.

Pros: Genuinely good skills, active community, zero cost, and you learn the internals by using it. Cons: You assemble and maintain it; no single-command installer or paid support; breadth is narrower than a full kit. Verdict (86/100): The pick if you want free and are willing to do a little plumbing. For most solo builders, this is enough. If you only clone one free thing from this list, make it this - it is the closest a free repo gets to feeling like a curated kit.

anthropics/skills · trusted official baseline

What it is: Anthropic's own repository of example and reference skills (github.com/anthropics/skills, accessed 2026-08-09). Best for: Anyone who wants first-party, vetted skills as a foundation before adding third-party ones. Price: Free, official.

Pros: First-party quality, safe defaults, great for learning how a skill is structured. Cons: Baseline coverage - it is a starting point, not a complete daily toolkit; you will add more. Verdict (84/100): Install this first, no matter what else you choose.

pr-review-toolkit · best free review plugin

What it is: A marketplace plugin focused on pull-request and code review. Best for: Teams and solo devs who want structured, repeatable review inside Claude Code. Price: Free (Anthropic marketplace).

Pros: One-command install, tightly scoped, does its one job well. Cons: Single-purpose - review only, not a general kit. Verdict (82/100): The best free answer if code review is your bottleneck.

feature-dev · best free feature-building plugin

What it is: A marketplace plugin that scaffolds and drives end-to-end feature development. Best for: Devs shipping features who want a guided build loop. Price: Free (marketplace).

Pros: Clean install, opinionated workflow, pairs well with pr-review-toolkit. Cons: Focused on the build loop; not broad coverage. Verdict (80/100): Grab it alongside #4 and you have a solid free build-and-review pair.

VoltAgent/awesome-claude-code-subagents · best subagent library

What it is: A large open collection of ready-made subagents for Claude Code. Best for: People who want specialist subagents (reviewer, tester, planner) without writing them. Price: Free, open source.

Pros: Big menu of subagents, easy to copy the ones you need. Cons: Quality varies across contributions; you cherry-pick and test. Verdict (78/100): A great parts bin, not a finished machine.

rohitg00/awesome-claude-code-toolkit · best for discovery

What it is: A curated "awesome list" pointing to tools, skills, and resources. (Note the name-twin: this is a discovery list, not the AgentKit kit and not OpenAI AgentKit.) Best for: Finding what exists before you commit. Price: Free.

Pros: Fast way to survey the ecosystem. Cons: It is a directory of links, not an installable kit - no curation of quality. Verdict (72/100): Use it to explore, then install from the entries above.

Official Anthropic plugin marketplace · vetted starting point

What it is: Anthropic's own plugin marketplace, where pr-review-toolkit, feature-dev, and github plugins live. Best for: Anyone who wants first-party-adjacent, one-command installs. Price: Free.

Pros: Trusted source, trivial install, good hygiene. Cons: Curated but limited catalog; you will still add community skills for full coverage. Verdict (76/100): The safe default for everyone before going third-party. Think of it as the app store rather than a specific app - the place you install #4 and #5 from, and the first place to check when a new first-party plugin lands.

Honorable mention: the scriptbyai Claude Code resource list is the most exhaustive directory out there (500+ projects), but it is a dump, not a ranking - no curation, no verdicts. Useful as a reference, overwhelming as a starting point.

Free vs paid - which should you actually install?

The honest answer depends on who you are. Here is the framework I would give a friend:

• Solo learner / hobbyist → go free. Install anthropics/skills plus obra/superpowers, add pr-review-toolkit and feature-dev from the marketplace, and you have a capable setup for $0. If you are still learning what skills do, paying for a kit is premature. Free is genuinely enough here.
• Shipping-daily indie / professional → a paid kit pays off. If you open Claude Code every workday, the hours you spend finding, wiring, and maintaining free repos are worth more than $99. AgentKit's breadth (108+ skills, 45 agents) and 5-minute setup remove that overhead, and lifetime updates mean you are not re-plumbing every month. This is where a kit earns its price.
• Team → paid + standardization. The real win for teams is not any single skill - it is everyone running the same curated set, so reviews and workflows are consistent. A maintained kit is easier to standardize on than "clone these seven repos and configure them yourself."

A rough way to run the math: if a paid kit saves you even one hour a month versus maintaining free repos, and you value your time at anything above $8/hour, the $99 has paid for itself inside a year - and lifetime updates mean you are not paying again. That is not a hard sell; it is arithmetic that only works if you are already a daily user. For a light or occasional user, the same math argues the other way, and free is objectively the right answer.

The failure mode to avoid: buying a paid kit before you have a real Claude Code habit, or refusing to pay $99 while burning ten billable hours a month gluing free repos together. Match the tool to your usage, not your pride. For a head-to-head on this exact question, see AgentKit vs free alternatives, and if cost is the sticking point, whether AgentKit is worth it.

Skip the assembly? If you already use Claude Code daily and just want breadth without maintaining a pile of repos, the paid kit is the shortcut. Check AgentKit (20% off via link).

AgentKit at a glance (why it's #1)

The verified facts behind the #1 ranking, straight from the site on 2026-08-09:

What tips it to #1 is coverage plus setup speed. The ak installer detects your OS and architecture, downloads the stable build, verifies a SHA-256 checksum, and drops the binary in ~/.local/bin - you authenticate with an email or API key and you are running:

# install ak, then
ak list skills # see everything installed
ak run <skill> # run a skill inside Claude Code

AgentKit is the successor to ClaudeKit, so if you used the old ck tool you can carry your setup over with ak migrate --from=ck (preview by default). More on that in the ClaudeKit to AgentKit migration guide.

The honest cons, restated: it is priced in USD and costs real money; you must already have Claude Code; and a slice of its skills overlap free repos, so if you only need review or one narrow workflow, a free plugin is the smarter buy. AgentKit wins when you want breadth, maintained - not when you need one tool.

FAQ

Is AgentKit the same as OpenAI AgentKit?

No. OpenAI AgentKit (Agent Builder, ChatKit, Connector Registry, launched October 6, 2025) is a platform for building your own agents. The AgentKit ranked #1 here is a curated kit for Claude Code at agentkit.best, installed with the ak CLI. Same name, different products.

Are Claude Code kits free?

Most are. The best free options are anthropics/skills, obra/superpowers, and marketplace plugins like pr-review-toolkit and feature-dev. Paid kits like AgentKit ($99 per kit, $149 bundle) sell curation, breadth, and maintenance on top of what you could assemble for free.

Do I need Claude Code Pro to use a kit?

You need Claude Code, and any paid Claude Code plan (Pro at $20/month or higher) works. Free repos and paid kits both extend Claude Code - they do not replace your subscription, so factor that cost in first.

Is AgentKit worth $99 versus free plugins?

If you use Claude Code daily, usually yes - the time saved on finding, wiring, and maintaining free repos outweighs $99, and updates are included for life. If you are a hobbyist or only need one narrow workflow, free plugins are the better call. It is a usage question, not a quality one.

Can I use kits with Cursor or Copilot?

Claude Code skill packs and marketplace plugins are built for Claude Code specifically. AgentKit positions itself to work across Claude Code, Codex, and Copilot CLIs, but always check current compatibility on the site before relying on it with a non-Claude tool.

How do I install a kit?

For free skill packs, clone the repo into .claude/skills/. For marketplace plugins, use Claude Code's plugin install command. For AgentKit, run the ak installer (native binary, no Node needed), authenticate, and it is ready in under five minutes.

Verdict & how to start

For 2026, start free: install anthropics/skills, add obra/superpowers, then pr-review-toolkit and feature-dev from the marketplace. That covers most solo work at $0. The moment Claude Code becomes a daily tool and maintaining repos becomes a chore, move up to a paid kit - AgentKit is the widest-coverage, fastest-to-set-up option, which is why it tops this list. Compare it against your current stack with Claude Code vs Cursor vs Copilot before you commit.

Prices and star counts were verified on 2026-08-09 and can change.