AI Templates for Claude Code

A searchable library of ready-made Claude Code components — skills, sub-agents, slash commands, MCP servers, hooks, settings and project templates. Filter by type and category, see what is trending this week, open any component to read its source fetched live from GitHub, and copy a single npx command to install it into your project. Data comes from the open-source claude-code-templates project (MIT). Full guide →

10 kindsMIT

Skills

Packaged know-how Claude loads on demand for a specific kind of task.

⌘K

Loading the library…

Catalogue data from davila7/claude-code-templates (MIT). Each component keeps its original license and author attribution.

Build your Claude Code setup from proven components

What is this?

AI Templates is a search front-end for the open-source claude-code-templates catalogue (MIT): 1,900+ skills, agents, slash commands, MCP servers, hooks, settings and project templates. Find one, read its source, then copy a single `npx` command to install it into your project.

atdev.blog only stores short descriptions for fast browsing. Each component’s full content is fetched live from GitHub when you open it, so you always see the original author’s latest version.

Which type for which job?

Skill: domain know-how Claude loads automatically when a task matches (e.g. writing tests, UI design). Agent: a sub-agent with its own role and tools that Claude can delegate to. Command: a slash command you trigger yourself for repeatable workflows.

MCP: connects Claude to external services such as databases, browsers or SaaS APIs. Hook: runs automatically before/after Claude edits files — ideal for formatting, linting or blocking risky commands. Settings: ready-made settings.json snippets for permissions, models and the status line.

Install safely

Run the install command from your project root; the CLI copies the component into `.claude/` (or the matching config file). Always read the source preview first — especially for hooks and MCPs, which can run shell commands or reach external services.

A few components ship under their own restrictive license: atdev.blog does not display their content and links to GitHub so you can read the terms yourself. After installing, commit the `.claude/` folder so your whole team shares the same setup.

Related Tools

All tools

Dev & PM Prompt Builder

Generate structured prompts and developer context for Claude Code and AI coding assistants

Text Utilities

Transform, organize, format, and clean text with powerful utilities

Hash Generator

Generate MD5, SHA-1, SHA-256, and other cryptographic hashes

Color Picker

Advanced color picker with HEX, RGB, HSL, and palette generation

About Claude Code Components, Ready to Install

Instead of writing every agent, skill or hook from scratch, start from a community-tested component. This library indexes the open-source claude-code-templates catalogue so you can compare options by popularity, read the actual source before installing, and add it to your .claude/ folder with one command.

Key Features

-1,900+ components across 10 types: skills, agents, commands, MCPs, hooks, settings, loops, mods, templates and sandboxes
-Instant search across name, category and description (⌘K / Ctrl+K)
-Filter by category and sort by popularity or name
-Weekly trending list based on real install counts
-Source preview fetched live from GitHub before you install
-One-click copy of the exact npx install command

Use Cases

Bootstrap a Claude Code setup

Pick a handful of agents and commands that match your stack instead of writing them from a blank file.

Connect external tools

Find an MCP server config for the database, browser or SaaS API you want Claude to work with.

Automate guardrails

Install hooks that format, lint or block risky commands before and after Claude edits files.

Learn by example

Read how experienced users structure skills and sub-agent prompts, then adapt the patterns to your own team.

Frequently Asked Questions

Where does this data come from?▾

From the open-source davila7/claude-code-templates repository (MIT). The catalogue is synced periodically; each component keeps its original author and license.

Does the install command change my project?▾

Yes — run it in your project root and the CLI copies the component into the .claude/ folder (or the matching config file). Read the source preview first and only install what you trust.

Why can some components not be previewed?▾

A few components ship under their own restrictive license. For those, atdev.blog shows only the name and description and links to GitHub, where you can read the content and its license terms.

What is the difference between a skill, an agent and a command?▾

A skill is know-how Claude loads automatically when a task matches; an agent is a sub-agent with its own role and tools that Claude can delegate to; a command is a slash command you trigger yourself.

Tips & Best Practices

-Start with the Trending list — it reflects what people actually installed this week
-Combine the category filter with search to narrow a large type like Skills quickly
-Always skim the source preview of hooks and MCPs: they can run commands or reach external services
-Commit the generated .claude/ files so your whole team shares the same setup