Claude Code Templates (aitmpl.com)

Claude Code Templates is a catalog plus a command-line installer. You find a component on aitmpl.com, such as a code-reviewer agent or a pre-commit hook, and one npx command copies it into your project’s Claude Code setup. It is less a single product than a large, mixed library gathered from many sources.

What it does (plain English)

The installer is an npm package you run with npx, which downloads and runs it without a permanent install. It knows six kinds of component:

Component	What it adds	Example flag
Agents	Specialist subagents	--agent development-tools/code-reviewer
Commands	Custom slash commands	--command testing/generate-tests
MCPs	Configs for MCP servers	--mcp database/postgresql-integration
Settings	Claude Code configuration	--setting mcp/mcp-timeouts
Hooks	Automatic actions at set moments	--hook git/prevent-direct-push
Skills	Skill folders	--skill <category>/<name>

You can combine flags to set up a stack in one go. The README’s example installs a frontend developer agent, a test-generation command and a GitHub MCP config together, with --yes to skip prompts.

In the repository copy we checked on 2026-10-18, the cli-tool/components folder held 424 agent files across categories like development-team, security, ui-analysis and podcast-creator-team, and 915 SKILL.md files. The package was at version 1.30.0.

The tool also has extra modes that are not about installing: --analytics for session monitoring, --chats for a mobile-friendly view of Claude’s responses (with an optional Cloudflare tunnel for remote access), --health-check and --plugins for a plugin dashboard.

Is it for you?

Useful if you like to browse visually and pick components one at a time, or you want a quick way to add a hook or setting you would otherwise write by hand.

Less useful if you want everything to stay updatable. Components are copied into your project, so they do not update themselves the way plugins from a marketplace can. And because the catalog pulls from many authors, quality varies; check each component before relying on it.

What you can do with it

1. Add a reviewer agent. npx claude-code-templates@latest --agent development-tools/code-reviewer --yes, then ask Claude Code to “use the code-reviewer agent on my last change.”
2. Add a test command. npx claude-code-templates@latest --command testing/generate-tests --yes, then run the new slash command.
3. Add a hook. npx claude-code-templates@latest --hook git/prevent-direct-push --yes adds the repo’s hook of that name. The README’s own hook example, git/pre-commit-validation, was not in the repository’s hooks/git folder when we checked, and two other README paths had moved (the MCP timeout setting now sits under settings/mcp/, the GitHub MCP config under mcps/integration/). Confirm a component path on aitmpl.com before you run a command.
4. Explore interactively. npx claude-code-templates@latest opens a menu to browse and install.
5. Check your setup. npx claude-code-templates@latest --health-check runs diagnostics on your Claude Code install.

How to set it up

1. Make sure Node.js is installed (node --version in a terminal).
2. Open a terminal in the project where you use Claude Code.
3. Find a component on aitmpl.com and copy its command, or use one of the examples above.
4. Run it. The --yes flag accepts defaults without asking.
5. Start or restart Claude Code in that folder. New agent folders need a restart the first time, per Anthropic’s subagent docs.

We did not run the installer for this profile. Running an npx package executes its code on your machine, and we limited our checks to reading the repository.

Watch out for

• It runs code you download. npx ...@latest fetches the newest version every time. Pin a version if you want repeatable installs, and read what a hook or MCP config does before adding it. Anthropic’s plugin security page notes that hooks run shell commands with your full user permissions.
• Anthropic’s document skills inside. The README credits 21 skills from anthropics/skills. The repository includes copies of the docx, pdf, pptx and xlsx skill folders (13 folders in all, under document-processing and scientific) carrying Anthropic’s proprietary LICENSE.txt, which forbids redistributing those materials. Use the built-in versions in Claude, or Anthropic’s own document-skills plugin, instead.
• Sponsored components. The README opens with a sponsor’s install command. That is advertising, not a recommendation from us or Anthropic.
• Remote chat viewer. --chats --tunnel exposes a view of your Claude conversations over the internet through Cloudflare. Use it only if you understand who can reach that link.
• Mixed licenses. The project is MIT, but bundled components keep their original licenses, listed in the README’s attribution section.

Alternatives

• VoltAgent subagents: a focused set of subagent files, installable as plugins.
• wshobson/agents: a plugin marketplace that keeps installs updatable.
• Claude Code official marketplace: Anthropic’s catalog with a security trust warning and context-cost estimates per plugin.