MCP Servers
MCP Servers
Many sample servers are usable right away; some technical knowledge is needed.
A repository of reference MCP servers and a list of community-built servers.
Overview
MCP Servers is the official GitHub repository containing reference MCP servers, used to illustrate MCP features and SDK usage. It is a good place to understand what an MCP server looks like before building your own.
The reference servers currently maintained by the MCP steering group are: Everything (a test server with prompts, resources and tools), Fetch (retrieve and convert web content), Filesystem (file operations with access control), Git, Memory (knowledge-graph-based memory), Sequential Thinking and Time. Many older servers such as GitHub, Slack, PostgreSQL and Google Drive have been moved to a separate archive repository.
The repository states clearly that these are educational examples, not production-ready solutions. To find published community servers, the repository points users to the MCP Registry.
Highlights
Official reference servers
Maintained by the MCP steering group, showing the standard-compliant way of doing things.
Filesystem with access control
Lets AI read and write files within configured directories.
Knowledge-graph memory
Helps AI retain information durably across sessions.
Fetch and Git
Fetches web content in a compact form for the model to read, and reads, searches and operates on Git repositories.
Quick install with npx or uvx
TypeScript servers run via npx; Python servers are recommended to run via uvx.
Configuration examples
The README has sample configuration for Claude Desktop, including how to run on Windows.
What it's for
Learn how to build a server
Developers read the Everything or Filesystem source to understand the structure before writing their own server.
Let AI work with documents
Office staff grant access to a reports folder so AI can summarize and organize files.
Remember customer information
Use the Memory server so an AI assistant keeps project context across many work sessions.
Support engineering teams
The Git server helps AI read history and search the company's code repositories.
Quick MCP experiments
Use the Time or Fetch server to demo MCP to leadership in a few minutes.
How to use it
- 1
Choose a suitable server
See the list of reference servers in the README and pick the one you need.
- 2
Run a TypeScript server
For servers written in TypeScript, run directly via npx.
$npx -y @modelcontextprotocol/server-memory - 3
Run a Python server
For Python servers, the README recommends using uvx.
$uvx mcp-server-git - 4
Or install with pip
You can install Python servers via pip and then run with python -m mcp_server_git.
$pip install mcp-server-git - 5
Register it in your AI app
Add the server to the mcpServers section of Claude Desktop's configuration file following the README sample.
Video guides
MCP Servers Explained & Built
Tech With Tim
you need to learn MCP RIGHT NOW!! (Model Context Protocol)
NetworkChuck
Tìm hiểu, xây dựng MCP Server và kết nối với Claude.ai, Agent - Mì AI
Mì AI
Real-world example
A small law office in Ho Chi Minh City wants to try an AI assistant on the secretary's computer. The IT person configures the Filesystem server to allow access only to the sample contracts folder, and the Memory server so AI remembers the office's file-naming conventions. The secretary asks AI to find and summarize clauses in the sample contracts without opening each file.
Tips & notes
Don't use directly in production
The repository states these are educational examples; add security measures yourself according to actual risk.
Limit directories for Filesystem
Grant only the folders you really need; avoid opening the whole drive.
Mind the Windows configuration
The README says to wrap npx with cmd /c when running on Windows.
Check archived servers
Servers such as GitHub, Slack and PostgreSQL have moved to the archive, so look for the new official version in the Registry.
FAQ
Are the GitHub and Slack servers still here?
No, they have been moved to the servers-archived repository. For example, Slack is now maintained by Zencoder.
What is the license?
New contributions are under the Apache License 2.0; existing code is under MIT.
Is there a server for Docker?
The README gives no Docker commands, only npx, uvx and pip instructions.
Where can I find more servers?
The README advises searching the MCP Registry for published community servers.
Useful links
Want to apply this in your business?
2dm looks at your workflows, picks the right tools and sets them up with your team. The first call is free.
Same topic
Model Context Protocol
4.5
The official MCP documentation site, an open standard for connecting AI assistants to external data and tools.
MCP specification
3.5
The full technical specification of the protocol: tools, resources, prompts and how clients and servers communicate.
MCP Registry
3.5
The official directory for finding and publishing MCP servers.