Claude Marketplaces & Plugins 2026: claudemarketplaces.com and the Plugin Ecosystem

Claude plugins are pre-packaged bundles of capabilities that drop into Claude Code (or Claude.ai) to extend functionality without you writing the integration code. The plugin ecosystem accelerated through 2025-2026 and now has official Anthropic-maintained plugins plus a growing community marketplace. This post is the practical tour.

What plugins are

A Claude plugin bundles together one or more of: MCP servers (tools the agent can call), skills (instruction packs), agent definitions (specialized sub-agents), hooks (lifecycle handlers), and configuration. Installing a plugin is one command; you get the full bundle wired up.

Think of plugins as "Claude Code apps" — the convenient unit of distribution for adding capabilities.

Plugins vs MCP vs Skills

MCP server — a single backend exposing tools.
Skill — an instruction pack (Markdown) Claude loads contextually.
Plugin — a packaged combination of MCP server(s), skills, agents, and configuration.

A plugin is the higher-level distribution unit. If you want "Stripe support in Claude," you install the Stripe plugin — which probably includes an MCP server (the API client), a skill (when to use what), and maybe some agent definitions (a "billing reviewer" agent). One install, capabilities wired in.

anthropics/claude-plugins-official

github.com/anthropics/claude-plugins-official is Anthropic's officially maintained plugin repository. Highest trust tier — signed by Anthropic, kept current with Claude Code releases.

Typical categories in the official set: code review, refactor agents, documentation generation, common API integrations (GitHub, Stripe, Slack, basic SaaS), data analysis pipelines. Browse the repo for the current catalog.

claudemarketplaces.com and the community

claudemarketplaces.com is a community-driven directory of Claude plugins. Categories, ratings, reviews, install metadata. The page claudemarketplaces.com/plugins/anthropics-claude-code indexes the official Anthropic plugins specifically.

Why it's useful:

Discovery — browse by category instead of searching GitHub.
Ratings — community feedback before you install.
Install instructions standardized across plugins.
Mix of official and community plugins.

Other resources in the same ecosystem: agentskills.io (skills directory), skillsmp.com (skills marketplace), various GitHub "awesome" lists.

Installing a plugin

The typical install flow in Claude Code:

# From inside Claude Code
/mcp install <plugin-name>

# Or via the plugins command if available
/plugins install <plugin-name>

# Verify
/mcp        # list MCP servers
/skills     # list active skills
/agents     # list available agents

Exact command varies by plugin and Claude Code version. Each plugin's README documents its specific install. The pattern is consistent: drop config into your project's .claude/ directory or use the relevant slash command.

Evaluating before you install

Plugins run with your Claude Code's permissions. A malicious plugin could read your files, exfiltrate data, or rack up API costs. Audit before installing:

Source. Official Anthropic plugins = safe. Well-known community contributors = likely safe. Anonymous one-time uploads = risky.
Permissions / tools. What does the plugin access? Network? Filesystem? Shell? Should match the plugin's purpose.
Code review. Plugin code is usually small enough to scan. Claude itself can audit a plugin's source for you — give it a 5-minute review.
Recent activity. Maintained recently (last 90 days) = better.
Stars / install count. Higher is loose validation, not a security proof.
Open issues for security concerns. A plugin with reports of "this exfiltrates X" should be a hard pass.

Plugins worth installing

For an iOS builder using Claude Code, useful categories include:

GitHub plugin — issues, PRs, code review without leaving the agent.
App Store Connect plugin — build status, metadata management, TestFlight from chat.
Railway plugin — deploy, logs, env vars from chat.
Code review plugin — structured agent that reviews your diffs.
Documentation generator plugin — produces API docs, READMEs.
iOS Simulator plugin — build, run, screenshot from chat.
Browser automation plugin — for tasks requiring a real browser.

Don't install everything — each plugin loads tools into Claude's context, and an over-equipped agent picks the wrong tool more often. Stick to ~5-8 frequently used plugins; install niche ones per-project as needed.

Authoring your own plugin

Roughly:

Write the MCP server(s) if your plugin includes tools.
Write the skill markdown if your plugin includes instructions.
Define agents if your plugin includes specialized sub-agents.
Bundle in a plugin manifest (format documented at code.claude.com/docs).
Publish to GitHub. Submit to claudemarketplaces.com for discovery.

For deeper coverage of the underlying primitives, see Claude Skills Ecosystem and Anthropic GitHub Repos.

See: Claude at Maximum Efficiency, Skills Ecosystem, Agentic AI in Claude, What is MCP?.

Sources & References