Search across 1047 pages

Try a tool name, category, or "lifetime deal"

Live directory · Updated weekly · No pay-to-rank

MCP Servers: The Complete Directory & Live Ranking

A complete directory of Model Context Protocol (MCP) servers, ranked by real adoption, GitHub growth, maintenance, and security, not opinion - with an in-depth page for every server.

228Servers ranked
501.9kCombined stars
12Categories
4Data sources
Live data from the official MCP Registry
Ranked by real GitHub adoption & growth
Security-checked via the GitHub Advisory DB
No server pays to be listed or ranked
See the full ranking methodology →
Updated July 6, 2026 Official MCP Registry + GitHub + npm/PyPI 228 servers ranked Refreshed weekly Independent & ad-free
What & how

What Is an MCP Server, and How Do You Install One?

What it is

An MCP server is a small program that exposes tools, data, and prompts to an AI application through the Model Context Protocol (MCP) - the open standard Anthropic released in November 2024. It runs locally over stdio, or remotely over Streamable HTTP, and any MCP-capable client can connect to it the same way, without custom integration code per app.

How to install one

Open your MCP client's config and point it at the server: edit claude_desktop_config.json in Claude Desktop, ~/.cursor/mcp.json in Cursor, .vscode/mcp.json in VS Code, or run claude mcp add in Claude Code. Every row in the directory below has a ready-to-copy install command for whichever client it supports - expand a row to get it.

Browse the full directory below ↓

As of July 6, 2026, this is a live, free MCP servers list ranking the best and most popular Model Context Protocol servers by real data. zplatform.ai tracks publicly listed MCP servers in the official MCP Registry that have a working install method and real adoption (228 ranked today, 501.9k combined GitHub stars). An MCP server is a program that exposes tools, data, and prompts to AI apps like Claude, Cursor, and ChatGPT through MCP - the open standard Anthropic released in November 2024. Ranked by adoption, growth, maintenance, and security, not opinion - and the list grows automatically as new servers are published. How we choose and rank →

The best MCP servers as of July 6, 2026, ranked by our composite score (adoption, growth, maintenance, trust and security):

  1. Codebase Memory - AI & Memory, 27.1k GitHub stars, quality score 96/100.
  2. Chrome DevTools MCP - Developer Tools, 46k GitHub stars, quality score 93/100.
  3. Agent Skills Search Server - Search & Web, 22.5k GitHub stars, quality score 92/100.
  4. mcp-server - Productivity, 42.6k GitHub stars, quality score 91/100.
  5. Oh My Posh Validator - Other, 23.0k GitHub stars, quality score 91/100.
Directory

Every MCP Server, Ranked

Every Model Context Protocol server that clears our quality gate and adoption floor, scored on a transparent composite of adoption, growth, maintenance, trust and security. Search, filter, or sort by any column - every row opens an in-depth profile with install snippets, security, and metrics.

228 servers
# Server Stars 30-day Downloads Updated Type Security Transport Quality
Editor's picks

Best MCP Servers by Use Case

Best MCP Servers for Local Memory & Context

Codebase Memory and basic-memory lead the AI & Memory category, giving assistants a persistent knowledge base or vector store to ground answers in your own documents instead of relying on training data alone.

Top Developer Tool MCPs

Chrome DevTools MCP and comfyui-mcp top the developer-tools category, the largest and most mature in the directory, wiring AI into code, repositories, and issue trackers.

Most Secure MCP Servers

Codebase Memory and Chrome DevTools MCP score highest on our security composite: no open advisories, and a clean assessment where a public homepage exists to check.

Best Free & Open-Source MCP Servers

Nearly every server in this directory is free to install. Codebase Memory and Chrome DevTools MCP are the highest-quality open-source picks - you only pay for any upstream service they connect to.

Best MCP Servers for Claude Code

Every server here works with Claude Code via a copy-paste "claude mcp add" command. Chrome DevTools MCP and Agent Skills Search Server are the top-scoring picks for coding workflows.

Browse

Browse by Category

Developer Tools

90 servers · 111.8k stars

Developer-tool MCP servers are the largest and most active category, wiring AI into code, repos, and issue trackers.

Filter the directory to Developer Tools ↓

Databases

6 servers · 5.2k stars

Database MCP servers let an AI query and inspect your data through a controlled interface.

Filter the directory to Databases ↓

Search & Web

37 servers · 69.2k stars

Search and web MCP servers give an AI fresh, real-world context beyond its training cutoff.

Filter the directory to Search & Web ↓

Design

8 servers · 22.4k stars

Design MCP servers connect AI to design tools and image generation.

Filter the directory to Design ↓

Productivity

14 servers · 56k stars

Productivity MCP servers connect AI to notes, docs, files, and project tools.

Filter the directory to Productivity ↓

DevOps & Monitoring

10 servers · 10.6k stars

DevOps and monitoring MCP servers expose infrastructure and observability data to AI.

Filter the directory to DevOps & Monitoring ↓

Communication

10 servers · 2.4k stars

Communication MCP servers let AI read and send across chat and email.

Filter the directory to Communication ↓

CRM & Sales

1 servers · 0 stars

CRM and sales MCP servers connect AI to customer and pipeline data.

Filter the directory to CRM & Sales ↓

Cloud

4 servers · 75.7k stars

Cloud MCP servers expose AWS, Azure, GCP, and edge platforms to AI.

Filter the directory to Cloud ↓

AI & Memory

18 servers · 38.1k stars

AI and memory MCP servers give assistants persistent memory and a knowledge base.

Filter the directory to AI & Memory ↓

Finance

10 servers · 8.9k stars

Finance MCP servers connect AI to payments, accounting, and market data.

Filter the directory to Finance ↓

Other

20 servers · 101.5k stars

Servers across every other niche are joining the MCP ecosystem.

Filter the directory to Other ↓
Data insights

State of the MCP Ecosystem

A live snapshot of 228 Model Context Protocol servers, built from the official MCP Registry, GitHub, npm/PyPI, and the GitHub Advisory Database. Refreshed weekly.

228Servers ranked
501.9kCombined GitHub stars
1.3MWeekly package downloads
93%Actively maintained
52%Official / first-party
53.1kCommunity forks

Servers by category

What MCP servers connect AI to, by count.

Adoption by GitHub stars

How server popularity is distributed.

Quality score distribution

zplatform.ai composite score (0 - 100).

Maintenance freshness

Time since the last commit or release.

Transport support

How clients connect (a server can support more than one).

Security posture

Advisory status via the GitHub Advisory Database.

Pricing

Free MCP Servers

Nearly every MCP server in this list is free and open-source - you install the server at no cost, and only pay for any upstream service it connects to (a database host, a SaaS API) using your own account. Use the "Free" filter in the directory to show only open-source, free-to-install servers.

Browse the MCP directory ↑
Get started

How to use these MCP servers

An MCP server runs as a small program (locally over stdio, or remotely over Streamable HTTP) that an MCP client connects to. To add one, open your client config and point it at the server: in Claude Desktop edit claude_desktop_config.json, in Cursor edit ~/.cursor/mcp.json, in VS Code add a .vscode/mcp.json, or run "claude mcp add" in Claude Code. Every row below has copy-paste install snippets for all four clients, plus the API keys it needs. Use the MCP Inspector to test a server before wiring it into your assistant.

Methodology

How we rank, and why you can trust it

  • Data sources: server identity, install methods, and status from the official MCP Registry; stars, releases, license, and last-commit recency from the GitHub REST API; weekly downloads from npm and PyPI; security advisories from the GitHub Advisory Database (GHSA). Refreshed weekly, with a daily snapshot for trend history.
  • Inclusion is automatic and transparent: we track every server in the official MCP Registry that has a description and a working install method (an npm, PyPI, OCI package or a public remote URL), excluding any marked "deleted" and badging any marked "deprecated". On top of that, a server must clear an adoption floor - at least 100 GitHub stars or 1,000 weekly package downloads - so the list stays credible rather than padded with auto-published stubs. We do not hand-pick winners; the ranking is driven by the data.
  • Adoption is a composite proxy, stated plainly: unlike a single canonical install count, MCP has no universal usage metric, so we use GitHub stars plus package downloads as the best free proxies for real adoption - never as exact install figures.
  • The 30-day growth trend is measured from our own weekly star snapshots, so it only appears once enough history has accumulated; until then a server shows no trend rather than a fabricated number.
  • "Official" means first-party: a server published under a namespace owned by the product vendor (a verified reverse-DNS domain, or a GitHub-organization namespace), as opposed to a community or aggregator re-publish. Every registry namespace is ownership-verified at publish time; the badge distinguishes vendor-built servers from community ones, and a short manual list corrects edge cases.
  • Security is assessed by mapping each server's npm/PyPI package to the GitHub Advisory Database. A server with no queryable package (remote-only or container-only) shows "not checked" rather than a false all-clear.
  • Ranking is data-driven only. No server pays to be listed or ranked, and there are no affiliate links influencing order.

The Quality Score formula

The Quality Score (0-100) is a transparent weighted blend: 35% Adoption (GitHub stars and weekly package downloads, log-scaled), 15% Growth (the 30-day star trend), 25% Maintenance (how recently the server was updated and its release cadence), 15% Trust (first-party namespace, an open-source license, and a public repository), and 10% Security (whether it has known, unpatched advisories). Raw popularity feeds Adoption but does not dominate the score.

FAQ

Frequently asked questions

What is an MCP server?

An MCP server is a program that exposes tools, data, and prompts to AI applications through the Model Context Protocol. It acts as a standardized bridge so an assistant like Claude, Cursor, or ChatGPT can read from and act on an external system - a database, a GitHub repo, a Slack workspace - without custom integration code.

What does MCP stand for?

MCP stands for Model Context Protocol, an open standard introduced by Anthropic in November 2024 for connecting AI applications to external tools and data sources.

How does an MCP server work?

An MCP host (the AI app) runs a client that connects to one or more servers over JSON-RPC. Servers expose three primitives - tools (actions the AI can call), resources (data it can read), and prompts (reusable templates). The two transports are stdio for local servers and Streamable HTTP for remote ones.

What are the best MCP servers?

The best MCP servers by our composite data score are listed in the leaderboard above, ranked on adoption, growth, maintenance, and security rather than opinion. Use the filters to narrow by category, transport, or whether a server is official and actively maintained.

Where can I find a list of MCP servers?

This page is a live MCP servers list built from the official MCP Registry, refreshed weekly. The canonical source is the official registry at registry.modelcontextprotocol.io; this leaderboard adds adoption, maintenance, and security data on top so you can compare and rank them.

Are MCP servers free?

Most MCP servers are free and open-source - you install them at no cost. Some connect to paid third-party services (a database host, a SaaS API) that require your own account and API key, but the server software itself is typically free. Each row shows the install method and required keys.

What are the most popular MCP servers?

The most popular MCP servers by GitHub stars and package downloads are shown in the "Most Popular MCP Servers This Month" view above. Popular servers cluster around developer tools (GitHub, filesystem), search and browser automation (Playwright), and databases (Postgres).

What's the difference between an MCP server and an API?

An API is a service-specific interface; every integration is bespoke. An MCP server wraps a system (which may itself sit on an API) in the standard Model Context Protocol, so any MCP-capable AI client can use it the same way. MCP is the universal adapter; an API is the underlying outlet.

MCP vs RAG - what's the difference?

RAG (retrieval-augmented generation) retrieves relevant text and injects it into a prompt. MCP is a connection protocol that lets an AI call live tools and read data sources directly, including taking actions. They are complementary: an MCP server can power retrieval for RAG, but MCP also covers actions that RAG alone does not.

What are the best MCP servers for Claude Code?

Every server in this leaderboard works with Claude Code, and each row includes a copy-paste "claude mcp add" command. For coding specifically, the most useful are the developer-tool, filesystem, database, and browser/search servers - GitHub, filesystem, Postgres, Playwright and similar. Filter the leaderboard to Developer Tools or sort by quality to compare the top picks for Claude Code.

Which AI tools support MCP servers?

Claude (Desktop and Claude Code), Cursor, VS Code (via GitHub Copilot), Windsurf, and a growing list of others support MCP servers. ChatGPT and Google Gemini have also added MCP support. Every server in this list ships install snippets for Claude Desktop, Cursor, VS Code, and Claude Code.

What is the official MCP Registry?

The official MCP Registry (registry.modelcontextprotocol.io) is the canonical, community-maintained index of MCP servers, where publishers register a server under an ownership-verified reverse-DNS namespace. This leaderboard is built from that registry, enriched with GitHub, npm/PyPI, and security data.

How do I install an MCP server in Claude Desktop?

Open Claude Desktop settings, edit claude_desktop_config.json, and add the server under "mcpServers" with its command, args, and any required env vars - then restart Claude Desktop. Every row in the leaderboard above provides the exact JSON to copy for Claude Desktop (and Cursor, VS Code, and Claude Code).

How many MCP servers are there?

The official MCP Registry lists thousands of published servers and grows daily. This leaderboard ranks the 228 that clear our quality gate and adoption floor (a description, a working install method, and at least 100 GitHub stars or 1,000 weekly downloads), as of July 6, 2026.

Who created MCP?

Anthropic created the Model Context Protocol and released it as an open standard in November 2024. It has since been adopted by many AI vendors and an open-source community that maintains the specification and the official registry.

Last updated: . Data from the official MCP Registry, GitHub, npm/PyPI, and the GitHub Advisory Database. This directory ranks servers by real data, not opinion - see the methodology and the machine-readable data.json.