Using a Claude Code Plugin

Mon, 01 Jan 0001 00:00:00 +0000

A Claude Code plugin is a directory that the harness can install, namespace, version, and reload as a unit. It is the packaged form of the same building blocks the harness already loads from .claude/: slash commands, skills, subagents, hooks, MCP servers, and a handful of less-common pieces (LSP servers, background monitors, on-PATH binaries, default settings). The difference is one of distribution and lifecycle rather than capability — anything a plugin does you could also do by hand in a project’s .claude/. The reason to use a plugin is that you don’t want to.

Authoring a Plugin

Mon, 01 Jan 0001 00:00:00 +0000

The first plugin most people author is for themselves: a few slash commands they keep re-creating in every new project, a hook they want everywhere, a subagent that has earned a permanent slot. The instinct is to keep that in .claude/ because there is only one user. That instinct is usually right for project-scoped knowledge and usually wrong for capabilities that travel — and the reason has nothing to do with sharing.

AI Plugins on Bitsy Services Wiki

Using a Claude Code Plugin

Authoring a Plugin