MCP Server List: The Complete Guide (2026)
Table of Contents
Finding the right MCP server used to mean digging through a GitHub README that kept changing. In early 2025 that README listed dozens of first-party and community servers in one long table. By April 2026 the official repo had retired that table entirely, pointing everyone to a dedicated registry at registry.modelcontextprotocol.io. The ecosystem had simply outgrown a single file.
This guide cuts through the noise. You will find a clear explanation of what an MCP server actually is, the current official reference servers (verified against the repo as of June 2026), the most useful third-party servers by category, where to discover more, and how to wire one up in Claude or ChatGPT. If you run WordPress, there is a dedicated section for that.
If you want a broader primer on the protocol itself, start with our complete guide to what MCP is, then come back here for the server list.
What is an MCP server?
An MCP server is any process that exposes capabilities to an AI client using the Model Context Protocol — the open standard Anthropic released on November 25, 2024 and subsequently donated to the Linux Foundation’s Agentic AI Foundation in December 2025. The protocol uses JSON-RPC 2.0 and defines three primitives a server can offer: Tools (functions the AI can call), Resources (data the AI can read), and Prompts (reusable prompt templates).
A server can run locally over STDIO — a subprocess the AI client spawns on your machine — or remotely over Streamable HTTP, meaning it lives on a server somewhere and the client connects to it over the network. Local servers are simpler to set up; remote servers unlock multi-user and headless scenarios.
The AI client (Claude, Cursor, n8n, etc.) discovers what a server can do by asking it at runtime. That is why you can drop a new MCP server into Claude Desktop and it immediately knows what tools are available without any code changes on the client side.
For a deeper dive, see our full MCP explainer.
Official reference servers (active, June 2026)
The modelcontextprotocol/servers repo on GitHub contains servers maintained by the MCP team. As of June 2026, seven servers remain active in src/:
| Server | What it does | Transport |
|---|---|---|
| Fetch | Fetches web pages and converts them to clean text for LLM consumption | STDIO |
| Filesystem | Secure read/write file operations with configurable directory access controls | STDIO |
| Git | Read, search, and manipulate local Git repositories | STDIO |
| Memory | Persistent knowledge graph stored in a local JSONL file | STDIO |
| Sequential Thinking | Structured multi-step reasoning via a thought-sequence tool | STDIO |
| Time | Current time and timezone conversion | STDIO |
| Everything | Reference/test server demonstrating all three MCP primitives | STDIO |
These are the canonical examples of correct MCP server implementation and a good starting point when building your own. All seven are published as npm packages under the @modelcontextprotocol/ scope.
What happened to the old reference servers?
Several early reference servers — including GitHub, GitLab, Google Drive, PostgreSQL, Puppeteer, Brave Search, Slack, Redis, Sentry, and Google Maps — were archived in late 2025 and early 2026. They moved to a separate modelcontextprotocol/servers-archived repo. Some were superseded by official vendor-maintained replacements (e.g., Brave now maintains its own brave/brave-search-mcp-server; Slack integrations moved to Zencoder). Always check archive status before referencing these in production.
Notable MCP servers by category (third-party, 2026)
The third-party server list in the main GitHub repo was retired on April 14, 2026 in favour of the official registry. What follows is a curated set of well-known servers across key categories — confirm current status at the registry before use.
Developer tools
| Server | What it does |
|---|---|
GitHub MCP Server (github/github-mcp-server) | Official GitHub server — create PRs, read issues, manage repos via GitHub API |
| GitLab MCP | Repository and CI/CD management for GitLab projects |
Playwright MCP (microsoft/playwright-mcp) | Microsoft’s official browser automation server; replaces archived Puppeteer |
| Sentry MCP | Query and triage Sentry error issues |
Data and databases
| Server | What it does |
|---|---|
| PostgreSQL (archived reference / community forks) | Schema inspection and read-only SQL queries |
| SQLite (archived reference / community forks) | Local database interaction and business intelligence |
| Redis (archived reference / community forks) | Key-value store access |
| Neon MCP | Serverless Postgres — branch, query, and migrate Neon databases |
Productivity and communication
| Server | What it does |
|---|---|
| Slack (Zencoder-maintained fork) | Channel management and message sending |
| Google Drive (community) | File search and content access |
| Google Calendar MCP | Read and create calendar events |
| Linear MCP | Issue tracking and project management |
Web and search
| Server | What it does |
|---|---|
Brave Search MCP (brave/brave-search-mcp-server) | Web and local search via Brave’s API — official replacement for the archived reference version |
| Firecrawl MCP | Web scraping, crawling, and structured data extraction |
| Exa MCP | Semantic web search with full document retrieval |
WordPress and CMS
| Server | What it does |
|---|---|
| Easy MCP AI | Turns any WordPress site into a fully compliant remote MCP server — 242 tools, no Node.js required |
The WordPress category is covered in detail below.
Where to find more MCP servers
1. The official MCP Registry
registry.modelcontextprotocol.io is now the canonical directory, replacing the GitHub README table that was retired in April 2026. You can search by category, filter by transport type, and see install instructions for each server.
2. GitHub — modelcontextprotocol/servers
The GitHub repo still hosts the active reference implementations and links to the registry. The servers-archived repo preserves older implementations for reference.
3. Awesome-MCP lists
Community-maintained “awesome” lists aggregate hundreds of servers with tags for language, category, and compatibility. Search GitHub for awesome-mcp-servers to find current versions — the lists update frequently.
4. AI client marketplaces
Claude.ai, Cursor, and Windsurf all have built-in discovery UI where you can browse and install servers directly from within the application. This is the easiest path for non-technical users.
How to add an MCP server to an AI client
The steps differ slightly per client, but the pattern is consistent.
Claude Desktop (local STDIO server)
- Open
claude_desktop_config.json(macOS:~/Library/Application Support/Claude/, Windows:%APPDATA%\Claude\). - Add an entry under
mcpServers:
{
"mcpServers": {
"filesystem": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/allowed/dir"]
}
}
}- Restart Claude Desktop. The new tools appear automatically.
Claude.ai / Claude Code (remote Streamable HTTP server)
For remote MCP servers that expose an HTTPS endpoint, go to Settings → Integrations → Add MCP Server in Claude.ai and paste the server URL. Claude Code accepts remote servers via claude mcp add --transport http <name> <url>.
Cursor / Windsurf / Cline
These editors expose an MCP configuration panel in their settings. You enter the server command (STDIO) or URL (HTTP), and the editor handles the rest. Refer to each editor’s documentation for the current UI location, as it changes between versions.
n8n
Add an MCP Client node to your workflow, point it at your server’s Streamable HTTP endpoint, and n8n will enumerate available tools automatically. This is particularly powerful for automating WordPress content workflows.
For a detailed walkthrough of connecting MCP to WordPress specifically, see our guide on Claude and WordPress via MCP.
Easy MCP AI: the WordPress MCP server
WordPress powers around 42% of the web, but until recently there was no clean way to give an AI agent real read/write access to a WordPress site. Existing approaches required running a separate Node.js proxy, maintaining a custom REST wrapper, or handing the AI raw API credentials with no tool-level scoping.
Easy MCP AI is a free, open-source WordPress plugin that solves this directly. Install it like any plugin and your WordPress site becomes a fully compliant remote MCP server accessible over Streamable HTTP — no Node.js, no separate process, no proxy.
Key facts
| Detail | Value |
|---|---|
| Total tools | 242 (96 core · 92 plugin-specific · 54 data integrations) |
| Supported AI clients | 16: Claude.ai, Claude Desktop, Claude Code, Cursor, Cline, Gemini CLI, GitHub Copilot, Google Antigravity, LibreChat, Manus, n8n, Pydantic AI, Roo Code, Windsurf, Zed, ChatGPT |
| Authentication | OAuth 2.1 with PKCE; API tokens are SHA-256 hashed before storage |
| Authorization | Per-tool scoping + standard WordPress capability checks |
| Hosting | Self-hosted on your server — no SaaS intermediary |
| Cost | Free and open source |
What you can do
With Easy MCP AI connected to Claude or another client, you can instruct your AI to:
- Draft and publish posts in the correct category with the right author
- Query WooCommerce orders and update product inventory
- Manage users, roles, and capabilities
- Run Advanced Custom Fields queries across post types
- Pull plugin-specific data through the 92 plugin tools
- Analyse site content and generate structured reports — all through natural language
Security model
Easy MCP AI’s own infrastructure stays entirely on your server. OAuth 2.1 with PKCE handles client authentication, API tokens are SHA-256 hashed before storage, and every tool call is checked against WordPress capabilities before execution. You control which tools are exposed at the per-tool level. There is no SaaS intermediary sitting between your site and the AI client.
Key facts summary
- The MCP protocol was released November 25, 2024 and donated to the Linux Foundation’s Agentic AI Foundation on December 9, 2025.
- The official third-party server list in the GitHub repo was retired April 14, 2026 — use registry.modelcontextprotocol.io instead.
- Seven first-party reference servers remain active: Fetch, Filesystem, Git, Memory, Sequential Thinking, Time, Everything.
- Many early servers (GitHub, Postgres, Puppeteer, Slack, Brave Search, etc.) are now in the
servers-archivedrepo, with several succeeded by official vendor versions. - Easy MCP AI is a self-hosted WordPress plugin that turns a WP site into a native remote MCP server with 242 tools and support for 16 AI clients.
Conclusion
The MCP server ecosystem grew fast enough that the original single-file server list could not keep up. The official registry at registry.modelcontextprotocol.io is now the authoritative source, and the seven active reference servers remain the best starting point for understanding what a well-implemented MCP server looks like.
For WordPress specifically, Easy MCP AI is a strong choice: self-hosted, no external dependencies, 242 tools, and support for every major AI client.
Get Easy MCP AI from the WordPress plugin directory
Official sources
- Model Context Protocol official site: modelcontextprotocol.io
- MCP Server Registry: registry.modelcontextprotocol.io
- Active reference servers (GitHub): github.com/modelcontextprotocol/servers
- Archived reference servers (GitHub): github.com/modelcontextprotocol/servers-archived
- Brave Search MCP (official): github.com/brave/brave-search-mcp-server
- Microsoft Playwright MCP: github.com/microsoft/playwright-mcp
- Easy MCP AI plugin: wordpress.org/plugins/easy-mcp-ai/