Building a Simple MCP Time Server with FastMCP and Cursor

During a live stream about using Cursor to track tasks in a large project, the author encountered frequent errors where the LLM wrote incorrect dates, prompting the need for a reliable time source.

They resolved the issue by installing the mcp-server-time package ( pip install mcp-server-time ) and configuring Cursor to use the MCP Time Server.

To deepen the learning, the author built a custom MCP server using the fastMCP framework, which provides tools, resource management, prompt definitions, client session handling, and SSE support for real‑time communication.

Setup steps include creating a project directory, initializing a virtual environment, and installing FastMCP:

mkdir mix-server

cd mix-server

uv venv venv

source .venv/bin/activate

uv pip install fastmcp

The main server code ( main.py ) defines a FastMCP instance and a tool that returns the local current time:

import datetime
from mcp.server.fastmcp import FastMCP

mcp = FastMCP("mix_server")  # important line

@mcp.tool()
def get_local_current_time():
    now = datetime.datetime.now()
    return now.strftime("%Y-%m-%d@%H:%M:%S")

if __name__ == "__main__":
    mcp.run()

Running the server is done with:

fastmcp dev main.py

Next, the author edited the Cursor configuration file ~/.cursor/mcp.json to point to the new server:

{
  "mcpServers": {
    "mix_server": {
      "command": "uv",
      "args": [
        "--directory",
        "/path/to/your/local/main_py_dir",  # replace with your actual path
        "run",
        "main.py"
      ]
    }
  }
}

With the server running, using Cursor’s auto‑run mode and the prompt “显示一下本地当前时间” (show the local current time) returns the correct timestamp, confirming the integration works.

The article also shares a typical prompt used to update a work plan file with timestamps, demonstrating how the custom tool can automate start and completion time markings.

In conclusion, the author reflects on solving a real‑world pain point, gaining experience in MCP server development, and encourages readers to explore further extensions of AI programming assistants.

Project resources are provided:

• GitHub: https://github.com/punkpeye/fastmcp • Documentation: https://modelcontextprotocol.io/docs