Claude Code now lets you install plugins: packages that bundle slash commands, subagents, MCP servers and hooks to adapt your development environment to concrete needs. The news comes straight from Anthropic and was published on October 9, 2025, when the company announced plugins are entering public beta for Claude Code. (anthropic.com)
What are Claude Code plugins
Think of a plugin as a little box you turn on when you need it and turn off when you don't. Inside you can find shortcuts (slash commands), specialized agents, connectors to external data via the Model Context Protocol (MCP), and customization points that change Claude Code's behavior at key moments.
Who is this for? For developers who want to standardize flows, for teams that need reproducible environments, and for maintainers of open source projects who want to help others use their tools correctly. Anthropic describes plugins as a lightweight way to package and share those customizations. (anthropic.com)
How they install and how they work in practice
Installing plugins is as simple as using the internal command: /plugin inside Claude Code. Anthropic says the experience works both in the terminal and in Visual Studio Code, and that plugins can be turned on or off to control system prompt complexity. That makes it easy to enable focused capabilities without overloading the session. (anthropic.com)
A quick example from the official docs:
/plugin marketplace add anthropics/claude-code
/plugin install feature-dev
If you've ever shared dotfiles or a setup for development, the idea is similar: share configuration and ready-to-use tools.
Plugin marketplaces and discovery
You can not only create plugins: you can publish marketplaces. A marketplace is a repository (git, GitHub or a URL) that contains a .claude-plugin/marketplace.json file listing available plugins. With that anyone can search, add and allow installation from within Claude Code. Anthropic encourages communities and organizations to build their own marketplaces to distribute approved internal solutions or share best practices. (anthropic.com)
The company even mentions community examples: engineer Dan Ávila's marketplace and Seth Hobson's curated sub-agent collection as active resources where you can find anything from DevOps automations to documentation flows. (anthropic.com)
Concrete use cases
- Enforce standards: a team can run automatic hooks to check style, tests, or security on every PR.
- User support: maintainers can add
slash commandsthat guide new contributors on how to run the project. - Share workflows: from debugging to deployment pipelines, all packaged and reproducible.
- Connect internal tools: install MCP servers to expose secure data to agents without losing controls.
Can you imagine a startup where every new developer boots up with the same set of commands and agents ready to use? Less friction, more productivity.
What this means for teams and founders
This isn't just another feature: it's a way to turn tacit knowledge into reproducible artifacts. For tech leads it reduces the costly gatekeeping of know-how; for founders it speeds up onboarding and the replication of workflows that work.
Also, because it's modular you can control the surface of risk: enable only what you need and keep the model context cleaner when you're focused on specific tasks.
So what can you do if you want to try it?
- If you already use Claude Code, open your terminal or VS Code and type
/pluginto see the plugins menu. Anthropic notes the feature is in public beta. (anthropic.com) - Check the official docs to start building your own plugin or publishing a marketplace at docs.claude.com. (anthropic.com)
- Get inspired by community marketplaces to spot useful patterns before creating your own.
To wrap up: this lowers the barrier to sharing how you work with AI day to day. It's not magic — it's reproducibility applied to model-assisted development. Which piece of your workflow would you package and share first?
