MCP.so: 1000+ MCP services search and share community

General Introduction

MCP.so is a program that focuses on Model Context Protocol (MCP) is a community-driven platform. It helps users discover, share, and use MCP servers, and allows AI assistants such as Claude to easily connect to a variety of external data and tools. At its core, the site provides a clear catalog of information about MCP servers, covering a wide range of resources from filesystems to databases, APIs, and more. Users can find off-the-shelf tools here or submit links to their own MCP servers. Through this platform, developers can build AI applications faster, and non-developers can also easily extend AI functions. mcp.so is updated in time, with information sourced from official sources and the community, and is highly practical.

Function List

• Server Browsing and Searching: Provides a list of MCP servers with support for searching by category or keyword.
• Submit MCP Server: Users can submit their own server links to share with the community via GitHub.
• Detailed server information: Shows the function, purpose, and connection of each MCP server.
• Community Support: Links to Telegram and Discord for communication and Q&A.
• open source resource: Based on the GitHub repository, all content is open and transparent, and users can participate in contributing.

Using Help

MCP.so is very easy to use and does not require a complicated installation process. All you need to do is visit the website and you can start exploring and using it right away. Here is a detailed guide to help you get started quickly.

How to Browse and Find MCP Servers

1. Open your browser and enter the URL https://mcp.so/, go to the home page.
2. There is a server list on the home page that lists all the MCP servers currently included.
3. If you're looking for a server with a specific feature, you can use the search bar at the top of the page. For example, type "GitHub" to find MCP servers that integrate with GitHub.
4. Clicking on a server name will take you to the details page. This will show you what the server can do, such as "Connect to PostgreSQL databases" or "Read local files".
5. The details page will also provide links to GitHub repositories where you can click to view the source code or documentation.

How to submit your own MCP server

1. Find the "Submit" button in the navigation bar of the website and click on it to go to the submit page.
2. You will be directed to GitHub's chatmcp/mcp-directory Issues page of the repository.
3. Click on "New Issue" and follow the template to fill in your MCP server information.
   • Includes server name, feature description, and GitHub link.
   • For example, "Name: mcp-server-github, Function: manage GitHub account, Link: https://github.com/xxx".
4. After submission, the community will review it. After review, your server will appear in the MCP.so list.

How to use the MCP server

The specific use of the MCP server needs to be paired with an AI client that supports the MCP protocol, for example Claude Desktop. the following are the basic steps:

1. Preparing the environment::
   • Make sure you have tools that support MCP, such as Claude Desktop or 5ire.
   • Download and install these tools (refer to the tools' official website for details on how to install them).
2. Operations Server::
   • Find the server you need from MCP.so, e.g. "mcp-server-filesystem".
   • Click the GitHub link to download the code locally.
   • Open a terminal, go to the code directory, and run the command (usually npm install Install the dependencies and then npm start (Initiation).
   • Upon startup, the server generates a local address, such as http://localhost:3000The
3. Connecting AI Clients::
   • Open Claude Desktop and find the MCP Server option in Settings.
   • Enter the address of the server you are running on (e.g. http://localhost:3000), click Connect.
   • Once the connection is successful, Claude will be able to use the features of this server, such as reading local files or querying the database.
4. test function::
   • Enter commands in Claude, such as "List my files" or "Query GitHub project".
   • The AI will perform the task through the MCP server and return the results.

Featured Feature Operation: Connecting Claude and MCP

At its core, MCP.so is about making AI assistants more powerful. Take Connect Claude, for example:

• Let's say you choose "mcp-server-github":
  • Once you download and run this server, Claude will be able to access your GitHub data.
  • Type "List my GitHub repositories" in Claude and it will return the results via the MCP server.
• Another example is "mcp-server-postgres":
  • Once started, configure your database connection information (enter username, password, etc. in the server's configuration file).
  • Enter "Query my sales data" in Claude and you'll get the results directly from the database.
    This approach allows AI to move beyond built-in knowledge to real-time access to external data.

caveat

• When running the server, make sure the network is open and the local port is not occupied.
• If you encounter problems, you can join MCP.so's Telegram or Discord community to ask other users for help.
• When submitting a server, try to provide detailed instructions for others to use.

application scenario

1. Developers building AI apps
   • You're a programmer and you want AI to automatically analyze code repositories. You can use MCP.so to find a GitHub server, connect to it, and have Claude generate code reports for you.
2. Extending AI capabilities for non-technical users
   • You're not good at programming, but you want to let the AI manage your schedule. mcp.so provides a ready-made Google Calendar server, so you can connect it and use your voice to command the AI to add a schedule.
3. Enterprise Data Integration
   • The company wants AI to query internal databases.MCP.so's database servers can be deployed quickly, and once connected, the AI can answer business questions.

QA

1. What is MCP.so?
   • MCP.so is a website that collects and shares servers for MCP, a protocol that enables AI to connect to external tools and data.
2. Do I need programming knowledge?
   • Not necessarily. Browsing and using an off-the-shelf server does not require programming, but running or submitting a server requires some basic skills.
3. Is the MCP server secure?
   • Yes. The MCP protocol has built-in security so that each server manages its own permissions and does not need to share sensitive information with the AI provider.
4. How do I know which server is right for me?
   • Check the server description on MCP.so to see if the features it supports match your needs, such as file management or API calls.