Browse the full directory of Claude Code plugins — commands, agents, skills, MCP servers, and more.
Browse plugins →A data-driven guide to MCP server plugins for Claude Code, covering persistent memory, browser automation, live documentation, and all-in-one solutions from the 7,159 MCP servers in the ecosystem.
MCP (Model Context Protocol) servers are the mechanism through which Claude Code plugins connect to external data sources and tools. Unlike skills and commands—which shape how Claude thinks and responds—MCP servers give Claude direct access to APIs, databases, file systems, and services outside the immediate project directory.
Across the Claude plugin ecosystem's 32,018 active plugins and 282,356 components, there are 7,159 MCP servers. That makes MCP the fourth-largest component type after skills (161,541), commands (58,116), and agents (48,047). While MCP is smaller in raw numbers, the plugins that include MCP servers tend to solve problems no other component type can: persistent memory, browser automation, live documentation, and external service integration.
This guide covers the MCP plugins developers are actually installing and starring, organized by the problem they solve.
Claude Code's context resets between sessions. MCP-based memory plugins solve this by running local servers that store and retrieve knowledge across sessions using semantic search. Three plugins stand out in this space, each with a different architectural approach.
claude-mem (68,843 stars) preserves persistent memory across Claude Code sessions. It lets you recall past solutions via semantic search, generate project journey reports from development history, build AI-powered knowledge bases for specific topics, map codebases into refactor flowcharts, and orchestrate subagent implementation plans with verification and anti-pattern checks. The plugin ships with skills, hooks, and MCP components—the hooks automatically capture context, while the MCP server handles storage and retrieval.
A standalone MCP-only variant is also available if you want just the memory server without the additional hooks and skills.
Mem0 (52,060 stars) takes a different approach to the same problem. It adds persistent memory via the Mem0 platform, storing user data, past decisions, strategies, and session states so Claude can retrieve relevant context when starting new tasks. Mem0 includes both Python and TypeScript SDKs for programmatic access, and its MCP server enables semantic search across long-term memories. It also ships with skills and hooks for automatic memory capture.
The key difference between claude-mem and Mem0 is architecture: claude-mem is fully local, while Mem0 integrates with the Mem0 platform for shared memory across tools and teams. If you work solo and want everything on your machine, claude-mem is the simpler choice. If you need memory that persists across team members or different AI tools, Mem0 is worth evaluating.
mempalace (50,068 stars) is the most focused of the three—it's a pure MCP plugin with 19 MCP tools dedicated to building a searchable "memory palace." It runs a local server, uses auto-save hooks for capture, and provides RAG-enhanced recall across sessions. The guided setup makes it the easiest to get started with if persistent memory is your sole concern and you don't need the broader skill and hook integrations that claude-mem and Mem0 provide.
Playwright (15,906 stars, 50 installs in the past 7 days) is Anthropic's official browser automation plugin for Claude Code. It wraps the Playwright testing framework in an MCP server, giving Claude the ability to run end-to-end tests, interact with web pages by clicking elements and filling forms, capture screenshots, and generate traces and PDFs—all running locally.
This is a pure MCP plugin: the entire integration is through the MCP server, not through skills or commands. That means Claude gains browser capabilities through tool use rather than prompt instructions, giving it structured access to page state, DOM elements, and screenshot data.
For developers doing frontend work, Playwright is one of the most practical MCP plugins in the ecosystem. Instead of switching between Claude and your browser to verify UI changes, Claude can navigate to the page, interact with elements, and screenshot the result within a single coding session. It's especially useful for writing and running end-to-end tests without leaving the Claude Code workflow.
Context7 (51,792 stars) solves a persistent problem in AI-assisted coding: Claude's training data doesn't always reflect the latest API changes for rapidly evolving libraries like React, Next.js, Vue, Prisma, or Supabase. Context7's MCP server fetches up-to-date, version-specific documentation, API references, and code examples directly into your Claude Code session.
You can query it via /context7:docs <library> [query] or use specific document IDs for precise lookups on setup, usage, and APIs. The plugin also includes commands, agents, and skills alongside its MCP server—the MCP provides the data pipeline to the documentation source, while the other components give you convenient interfaces for access.
If you frequently work with libraries that ship breaking changes between versions, Context7 eliminates the guesswork of whether Claude is suggesting code for the right version.
prompts.chat (159,878 stars—the highest-starred plugin in this list) goes beyond documentation into prompt management. Its MCP server connects Claude to the prompts.chat library, letting you search, retrieve, improve, and manage thousands of AI prompts and Claude skills. You can install skills to extend capabilities, fill prompt variables, save custom prompts with metadata, and enhance them using AI.
With commands, agents, skills, and MCP components, prompts.chat is one of the more comprehensive plugins in the ecosystem. The MCP server acts as the bridge to the external prompt database, while the skills and commands handle the user-facing interaction layer. For teams standardizing on prompt patterns or developers who want a searchable library of proven prompts, it's a practical addition.
everything-claude-code (169,421 stars, 293 installs in the past 7 days) is the highest-installed MCP-equipped plugin by weekly velocity in this group. It orchestrates team-oriented AI agent workflows using 48 specialist agents, 183 skills, and 79 commands to plan, generate, review, test, and deploy code across 20+ tech stacks. Its hooks enforce quality checks, and the MCP server integrates external tooling into those workflows.
This plugin represents the bundled approach—rather than installing separate plugins for memory, testing, and code generation, everything-claude-code packages them into a single install. The tradeoff is scope: at 48 agents and 183 skills, it's significantly more opinionated about your workflow than a focused plugin like Playwright or mempalace. If you want comprehensive coverage and are willing to adopt its conventions, it delivers the broadest capability set. If you prefer composing smaller, focused tools, the individual plugins in this guide are a better fit.
When evaluating MCP plugins, consider three factors:
Scope of integration. Plugins like mempalace (MCP-only, 19 tools) are narrow and predictable. Plugins like everything-claude-code (commands, agents, skills, hooks, and MCP) reshape your entire workflow. Match the scope to your actual needs—broader isn't inherently better.
Local vs. external. claude-mem and mempalace run entirely locally. Mem0 connects to an external platform. Context7 fetches from public documentation APIs. Playwright runs a local browser instance. Know where your data flows, especially if you work with proprietary code.
Composability. MCP plugins can coexist. Running Playwright for browser testing, Context7 for documentation, and claude-mem for memory doesn't create conflicts—each MCP server handles a different domain. The all-in-one plugins are less composable since they cover overlapping ground.
The MCP servers directory lists all 7,159 MCP servers across the ecosystem. For developers evaluating which to install, the plugins above represent the highest-starred and most-installed options with MCP components as of April 2026—covering the core use cases where external integration adds genuine value to your Claude Code workflow.
Supercharge Claude Code with 300+ agents, skills, commands, and hooks to orchestrate autonomous multi-agent coding workflows, enforce TDD, conduct security audits, generate production code across JS/TS/Python/Rust/mobile stacks, optimize performance, and automate deployments/testing.
Automate browsers and run end-to-end tests with Playwright directly in Claude. Interact with web pages by clicking elements, filling forms, taking screenshots, generating traces, and executing testing workflows locally via npx subprocess.
Search, retrieve, improve, and manage thousands of AI prompts and Claude skills from prompts.chat directly in your coding assistant. Install skills to extend capabilities, fill prompt variables, save custom prompts with metadata, and enhance them using AI.
Memory compression system for Claude Code - persist context across sessions
Preserve full context across Claude Code sessions in a local SQLite database by capturing tool observations, session summaries, and codebase details. Query past work via semantic search, build project-specific AI knowledge bases, generate development history reports, map architectures for refactors, and orchestrate subagents with historical insights for phased, verified implementations.
Add persistent memory to Claude Code tasks and AI apps via Mem0: retrieve relevant past decisions, strategies, and session states on new tasks; store user data for personalization; enable semantic search across long-term memories using Python/TS SDKs, hooks, and MCP tools.
Fetch up-to-date, version-specific documentation, API references, and code examples for libraries like React, Next.js, Vue, Prisma, and Supabase directly into your LLM context using Context7 skills, commands, agents, and MCP server. Query via /context7:docs <library> [query] or IDs for precise lookups on setup, usage, and APIs.
Run a local MCP server with mempalace-mcp to give AI persistent memory: mine projects and conversations into a searchable palace using auto-save hooks, 19 MCP tools, and guided setup for RAG-enhanced recall across sessions.