MCP Reference Servers

MCP Reference Servers gives developers examples for connecting AI applications to tools and data. Maintained by the MCP steering group, these are educational implementations, not production-ready servers.

• Files and repositories: Filesystem supports file operations with configurable access controls, while Git provides tools to read, search and modify repositories.
• Persistent memory: Memory stores information in a knowledge graph for use by a connected AI client.
• Web content and time: Fetch retrieves web content and converts it for LLM use; Time handles time and timezone conversions.
• Problem-solving and testing: Sequential Thinking supports reasoning through sequences of thoughts, while Everything is a test server for MCP prompts, resources and tools.

The collection includes TypeScript and Python servers that run locally and need an MCP client, such as Claude Desktop, to use their capabilities. ChatGPT, Visual Studio Code and Cursor also support the protocol. Fetch accesses external websites.

Developers adapting these examples for production need to assess their security requirements and add safeguards appropriate to their use case. The MCP Registry is a separate directory for published servers.