There are dozens of posts about variations of Cline’s Memory Bank ( Cline Memory Bank | Cline ). Most of them do an excellent job - context files that describe different aspects of your project significantly improve the vibe-coding experience.
But there’s one catch: filling out all those context files manually for every single project can be tiring.
To solve this, I built a simple MCP server that helps to generate Memory Bank files locally: GitHub - ipospelov/mcp-memory-bank: MCP server that helps build Memory Bank - structured documentation system for context preservation
How it works:
-
Write a brief description of your project - no special format required
-
Ask Cursor to build a Memory Bank:
Create Memory Bank files with your tools based on *your_description*
-
Cursor fetches templates via the MCP server
-
It creates context files based on your description and the templates
-
As you keep working, Cursor updates the Memory Bank automatically
It is also important to move memory_bank_instructions.md into native Cursor rule with .mdc extension and apply it always.
You can also use it to generate a Memory Bank for your codebase. Just ask:
Analyze and describe project. Create Memory Bank files with your tools based on description
Here’s how to setup the MCP server in your Cursor mcp.json config:
{
“mcpServers”: {
“mcp-memory-bank”: {
“command”: “uvx”,
“args”: [
“–from”,
“git+https://github.com/ipospelov/mcp-memory-bank”,
“mcp_memory_bank”
]
}
}
}
I also created an interactive course that shows how to set up a Memory Bank and build applications with it. It works within Cursor IDE and guides you from setting up MCP Server to running an application.
Check it out here, it’s free: https://enlightby.ai/projects/37
Hope you find this useful!
