Intermediate
Claude Agents Explained: Subagents, Agent SDK, Managed
Claude agents come in four forms: Claude Code subagents, the Agent SDK, Managed Agents and the API tool runner. See how each works and which to pick.
Independent · Community-built · Tested and dated
Tested kits and guides to get more done with Claude.
Claudekit helps you get work done with Claude: ready-to-run Claude Code kits, step-by-step guides, Skills, MCP servers, prompts and free tools. Every kit shows how we tested it, and every fact carries the date we checked it.
Install our kits by name: add the Claudekit marketplace in Claude Code once.
/plugin marketplace add claudekit-ai/kits New here? Read the Claude kit overview or how we test.
Start with a goal
Pick the job in front of you. Each one starts with the most direct help we have.
-
Keep Claude Code away from dangerous commands
Block deletes you cannot undo and get asked before force-pushes, hard resets and dropped tables.
Start with this kit
Safe Mode -
Review a branch before you open the pull request
Get a written review with findings by severity and file references, offline from your git history.
Start with this kit
PR Review -
Write release notes from your commits
Turn git log into notes grouped by change type, breaking changes first, ready to paste.
Start with this kit
Release Notes -
Use fewer tokens and stay under your limits
See what loads into every session, watch your plan usage, and trim what you do not need.
Start with this kit
Token Diet -
Connect Claude to your tools with MCP
Find an MCP server by use case, from code and docs to browsers and data, then generate the config.
Start with this directory
MCP servers -
Extend Claude Code with plugins and Skills
Install plugins others already built, with the exact commands, and learn when a Skill fits better.
Start with this directory
Plugins -
Write a CLAUDE.md that Claude actually follows
Give Claude Code your stack, commands and rules in one short project file.
Start with this free tool
CLAUDE.md Generator -
Pick a plan or estimate your API bill
Compare your usage with plan prices, or price a workload per request, day and month.
Start with this free tool
Plan picker
Free tested kits
Small Claude Code plugins that do one job. Each badge says how far our testing went; see how we test.
-
A REVIEW.md in your repository with a verdict, findings by severity with path:line references, and a filled-in review checklist.
Tested with Claude Code 2.1.290 · 2 min setup
/plugin install pr-review@claudekit -
A RELEASE_NOTES.md with your changes grouped into breaking changes, features, fixes and the rest, ready to paste into a CHANGELOG.
Tested with Claude Code 2.1.290 · 2 min setup
/plugin install release-notes@claudekit -
Claude Code cannot run rm -rf on broad paths or pipe downloads into a shell, and asks you before it rewrites git history or drops data.
Tested with Claude Code 2.1.290 · 2 min setup
/plugin install safe-mode@claudekit -
You know which always-loaded files cost context on every request, and you have a short list of documented trims to apply.
Tested with Claude Code 2.1.290 · 2 min setup
/plugin install token-diet@claudekit -
Your Claude Code status line shows the model, context use and how much of your 5-hour and weekly plan limits you have used.
Tested with Claude Code 2.1.290 · 3 min setup
/plugin install usage-meter@claudekit
Installing by name needs the marketplace added once (the command at the top of this page). A kit can run hooks with your permissions, so read its files before you install it.
Free tools
They run in your browser, and nothing you type leaves your device.
- CLAUDE.md Generator Project memory for Claude Code: stack, commands, conventions and rules.
- SKILL.md Generator A spec-valid Skill with frontmatter, instructions and folder layout.
- MCP Config Generator claude mcp add commands, .mcp.json or Claude Desktop config.
- Claude API Cost Calculator API cost per request, day and month, plus a Pro, Max and Team plan picker.
- Plan picker Compare your Claude Code usage with Pro, Max and Team plan prices.
Explore the library
- Recipes 5 Claude Code kits we ran, with the test level, version and date.
- Guides 58 Step-by-step tutorials for Claude Code, the API, MCP and agents.
- Skills 30 Reusable Skills with install commands, sources and licenses.
- Prompts 11 Copy-ready prompt templates for current Claude models.
- Tools 73 MCP servers, CLIs, plugins, apps and starter kits.
- News 4 What changed this week across Claude and its ecosystem.
- Compare 11 Honest side-by-sides to help you choose faster.
- MCP servers MCP servers for Claude, grouped by what you want to connect.
Latest guides
View all → Beginner
AI Agent Development Cost (2026): Build, Run and Maintain
What a custom AI agent costs in 2026: real build prices from freelancers, fixed-price services and agencies, plus how to estimate token and upkeep costs.
Beginner
AI Agents for Small Business: 12 Use Cases, DIY or Hire
Twelve narrow jobs an AI agent can do in a small business, each with its trigger, tools, a feature count and whether you can build it yourself or should hire.
Latest tools
View all → Kit Featured
Agentkit.best (formerly Claudekit.cc): Claude Code Kit Review
Claudekit.cc changed to Agentkit.best: a paid kit of skills, subagents and slash-command workflows for Claude Code and other agents. What it includes, pros and cons.
CLI Featured
Claude Code: Anthropic's Agentic Coding CLI
Claude Code is Anthropic's agentic coding tool for the terminal, IDE, desktop and web. Install commands, use cases, pros and cons, and where to start.
★ 153,195
MCP Server
Apify MCP Server: Run Web Scrapers from Claude
Apify's official MCP server lets Claude find and run Actors from Apify Store to scrape sites, maps and social media. Claude Code plugin, OAuth, caveats.
★ 12,930
Latest news
View all →Claudekit.cc Is Now Agentkit.best: What Changed and Where to Go
The Claude Code starter kit Claudekit.cc changed its name to Agentkit.best and moved from claudekit.cc to agentkit.best. What that means for existing users.
Claude Code v2.1.309: Sandbox and Permission Fixes
Claude Code v2.1.309 tightens sandbox and Bash permission checks, protects MCP config files, and changes how sed -i and claude ultrareview behave.
Claude Models on Google Cloud (Anthropic Managed) Launches
Anthropic launched Claude Models on Google Cloud, the Claude API run by Anthropic via your Google Cloud project, with IAM access and Marketplace billing.
Community
Built something that helps? Share it.
Send us a kit, Skill, MCP server, plugin or prompt collection you made. Listings are free and we review every submission. Missing a goal above? Email us the job you want help with.
Need a custom AI agent? Hire Claudekit.
A custom AI agent for your business, priced by its number of features. Our AI agents build and test it, and you get a private repo.
Built in public: a team of AI agents and one human editor run this site.
The agents go live with the site. See the agents at work →
From the Claudekit workshop
Kiln Guard: guardrails for your coding agent
Blocks the commands you can’t undo, keeps secrets out of the chat, and runs your tests before Claude Code says “done”. Hooks, skills and subagents in one small plugin.
See what’s inside → Launching soon — get notified
$19 launch price, then $29. One-time payment · 14-day refund · Not on sale yet
Get the weekly Claude digest
New skills, MCP servers, prompts and Claude updates — one short email a week. No spam, unsubscribe anytime.
The newsletter is launching soon. Follow the RSS feed in the meantime.