DISCOVER THE BEST MCP SERVERS, CLIENTS, AND TOOLS

Curating the context layer of the Agentic Web

MCP Directory is a clean, developer-first hub for finding Model Context Protocol (MCP) servers, UI clients, and smart resources to empower context-grounded AI agency.

Official MCP Servers 9

Core endpoints provided by the protocol and foundation teams.

View all

Featured MCP Servers 3

Hand-reviewed developer stacks connecting services and APIs.

View all

Trending / Latest Servers 6

Highly active context gateways capabilities and resources.

View all

Compatible with Leading AI Clients

Seamlessly integrates with the most popular AI development environments and platforms.

Claude Code
Cursor
Zed
VS Code
Windsurf
Gemini CLI
ChatGPT
Goose
Codex
opencode
Cline

Plus support for any MCP-compatible AI tool through manual configuration

Frequently Asked Questions

Quick answers to help you configure, secure, and understand Model Context Protocol integrations.

The Model Context Protocol (MCP) is an open-source development standard created to allow AI models and large language agents to formally and securely connect to external APIs, databases, browser runtimes, and local developer file systems.
An MCP server is a lightweight microservice or local process that exposes specific resources, prompts, and tools via the standard protocol. It acts as an integration gateway, translating generic model requests into concrete database queries, shell executions, or third-party API actions.
MCP servers run locally or in the cloud. When an AI client (like Claude Desktop or Cursor) detects that an agent needs information or wants to perform an action, it leverages the server's declared JSON-RPC tools, retrieves the output, and feeds that structured context back to the model's chat window.
Usually, local servers are configured within your developer tool settings file. For example, in Claude Desktop, you specify the configuration in `claude_desktop_config.json` under your OS user path with the execution command (e.g. `npx -y @modelcontextprotocol/server-postgres`).
Because MCP servers run within your local system architecture (under your user's permissions), they pose standard safety boundaries. When installing community servers, inspect their official source code, check for authentication requirements, and sandbox permissions if they modify local files.
Yes! Click on 'Submit Server' in our navbar, fill out the markdown metadata template, and open a pull request. We review submissions for production-ready code guidelines, authentic tags, and documentation before deployment.