Customize Claude Code with plugins (www.anthropic.com)

Claude Code now supports plugins — lightweight packages that bundle slash commands, subagents (purpose-built agents), MCP (Model Context Protocol) servers, and hooks into Claude Code’s workflow so you can install a whole customization with a single /plugin command (public beta). Plugins are toggleable to keep system prompt context lean, work across the terminal and VS Code, and make it easy to share standardized dev environments, debugging or deployment workflows, testing harnesses, and integrations with internal tools and data sources. Technically, plugins expose four extension points (commands, agents, MCP servers, hooks), can be hosted and discovered via simple plugin marketplaces (a git/GitHub repo or URL containing a .claude-plugin/marketplace.json), and are installed/managed through the /plugin menu (e.g., /plugin marketplace add anthropics/claude-code; /plugin install feature-dev). For the AI/ML community this reduces onboarding friction, enforces team best practices (code-review/test hooks), and spurs ecosystem growth — community authors are already publishing marketplaces and hundreds of subagents. Expect faster experimentation, reproducible agent setups, and richer tool integrations, though teams should consider security and configuration controls when exposing internal MCP servers or hooks.