Back to directory

Claude Code Plugins Directory

Anthropic's official, curated marketplace of Claude Code plugins, listing internal plugins built by Anthropic alongside reviewed third-party plugins, installable directly from Claude Code.

Claude Code Plugins Directory

The directory splits its listings into internal plugins developed and maintained by Anthropic and external plugins submitted by third-party partners and the community, which must meet quality and security standards before acceptance. Each plugin follows a standard structure, with a metadata file, optional MCP server configuration, slash commands, agent definitions and skill definitions.

A plugin's marketplace name is an immutable slug once published, since users install under that slug; a separate displayName field can be changed freely, and a renames map lets a maintainer migrate existing installs if a slug must change. A plugin whose source repository ships bare skill files without a full plugin manifest can still be listed using a strict-mode declaration that points at each skill directory directly.

Plugins install with a single Claude Code command, and Anthropic notes it does not control or verify the contents of third-party plugins, so a plugin should be reviewed before use. It is written in Python and licensed under Apache-2.0.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.