Prompts

The Prompts MCP Server lets you manage a collection of prompt templates stored as markdown files with optional YAML frontmatter. It exposes a simple, file-based data store that you can access from any MCP client to add, retrieve, list, or delete prompts on the fly, with real-time caching for fast, up-to-date results.

How to use

Connect your MCP client to the Prompts MCP Server to manage prompts from your workflow. You can add new prompts, retrieve specific prompts by name, list all available prompts with metadata previews, and delete prompts when needed. Prompts are stored as markdown files in a dedicated directory and can include YAML frontmatter for structured metadata like title, description, tags, and difficulty. The server keeps an in-memory cache and watches the filesystem so changes are reflected immediately.

How to install

Prerequisites: you need Node.js 18 or higher installed on your system.

npm install -g prompts-mcp-server

Additional setup and usage notes

Configure your MCP client to connect to the server. You typically provide a command to launch the server so the client can communicate via MCP. Include an environment variable to specify where prompts are stored.

{
  "mcpServers": {
    "prompts-mcp-server": {
      "command": "prompts-mcp-server",
      "env": {
        "PROMPTS_FOLDER_PATH": "/path/to/your/prompts/directory"
      }
    }
  }
}

Using the server tools

The server provides a compact set of tools to manage prompts.

• add_prompt: Add a new prompt to the collection. If no YAML frontmatter is provided, default metadata is added automatically.

• create_structured_prompt: Create a prompt with guided metadata and validation for consistent structure.

• get_prompt: Retrieve a prompt by name.

• list_prompts: List all prompts with metadata previews.

• delete_prompt: Remove a prompt by name.

File storage and metadata

Prompts are stored as markdown files in a dedicated prompts/ directory. You can enable YAML frontmatter for structured metadata such as title, description, category, tags, and difficulty. The server monitors file changes in real time and updates the in-memory cache accordingly.

Starting and verifying the server

Start the server after installation and configuration to verify it runs without errors. You can also test with an MCP Inspector tool to ensure the client can connect successfully.

npm start

or verify with MCP Inspector:

npx @modelcontextprotocol/inspector prompts-mcp-server

Available tools

add_prompt

Add a new prompt to the collection. If no YAML frontmatter is provided, default metadata will be automatically added.

create_structured_prompt

Create a new prompt with guided metadata structure and validation.

get_prompt

Retrieve a prompt by name.

list_prompts

List all available prompts with metadata preview.

delete_prompt

Delete a prompt by name.