Personal MCP starter kit

Motivation

AI tooling is evolving rapidly, with vendors each adding their own methods of "configuring" their models - various different markdown files like CLAUDE.md, GEMINI.md and the slightly more generic AGENTS.md. These methods work well for consistent setup across multiple developers, for a single project. This repo explores the idea of encapsulating your individual workflow as a developer as a set of MCP utilities (mostly prompts), in a way that is portable between AI tools.

Overview

This repo contains a Go MCP server that includes all prompts in the prompts/ directory, embedded in the binary itself - no need to copy around other files!

Prompts require a bit of YAML frontmatter, a name attribute at a minimum, and the remainder of the document is used as the prompt text. There are a few examples provided you can use for inspiration.

Because the prompts are embedded in the binary, updating prompts requires you to rebuild the binary. This also means that your prompts can be stored in version control, and you can determine the version of the prompts from the build data included in the binary.

Usage

1. Clone this repo

2. Add or edit files in prompts to suit your use case

3. run go build . - you now have an MCP server

4. Configure your AI tools to access this MCP server. Configs vary, but it might look something like this:

   "mcpServers": {
       "personal": {
           "command": "/path/to/your/mcp-starter"
       }
   }
   

Your prompts are now available in your AI tool of choice. They can be shared across multiple tools, too!

Why MCP?

With so many available configuration methods, why did I choose MCP?

For me, the major reason is flexibility. AI tools are changing rapidly, with a new tool or model emerging every week. To remain agile and avoid lock-in, I want to be able to use the same tools in any AI system. Today, the "greatest common factor" among AI tools is MCP. The protocol may not be ideal, but it is supported everywhere, and is one of the few standards in a rapidly evolving space.