MCP Prompts Server

An MCP (Model Context Protocol) server for managing, storing, and providing prompts and prompt templates for LLM interactions.

Features

• Store and retrieve prompts and templates
• Apply variable substitution to prompt templates
• Tag-based organization and search
• MCP Prompts protocol support
• Tools for prompt management
• Import and process prompts from various sources
• Export prompts in different formats (JSON, Markdown)
• Share and import prompt collections
• Docker and devcontainer support
• Comprehensive test client
• New: PGAI database integration with semantic search
• New: High-quality professional prompt collection

Installation

Quick Install

The simplest way to install is using the unified installer script:

./install.sh

Installation Options

The installer supports multiple installation modes:

# Install locally (default)
./install.sh --mode=local

# Install globally with npm
./install.sh --mode=npm

# Build and install as a Docker image
./install.sh --mode=docker --docker-user=yourusername

# Process raw prompts during installation
./install.sh --process-prompts

# Install in development mode
./install.sh --dev

For more options, run:

./install.sh --help

Manual Installation

1. Clone this repository:

   git clone https://github.com/yourusername/mcp-prompts.git
   cd mcp-prompts

2. Install dependencies:

   npm install

3. Build the project:

   npm run build

Docker Installation

Build and run with Docker:

# Build the image
docker build -t yourusername/mcp-prompts .

# Run the container
docker run -it --rm yourusername/mcp-prompts

Usage

Running the Server

Run the server with:

npm start

Prompt Processing Pipeline

The easiest way to process, tag, and organize prompts is with the complete pipeline:

# Create your rawprompts.txt file first, then run:
npm run prompt:pipeline

This runs the complete process:

1. Extracts prompts from the raw file and generates metadata
2. Performs intelligent tagging based on content
3. Organizes prompts into appropriate category directories
4. Summarizes the results

Additional options:

# Preview without making changes
npm run prompt:pipeline:dry

# Show detailed output from each step
npm run prompt:pipeline:verbose

# Run pipeline but keep the raw prompts file
npm run prompt:pipeline:keep

Processing Raw Prompts

To process raw prompts from a file:

# First create a rawprompts.txt file with your prompts
npm run prompt:process

This will:

1. Extract prompts from the raw file
2. De-duplicate similar prompts
3. Generate metadata, tags, and descriptions
4. Export as both JSON and Markdown formats
5. Place them in the processed_prompts directory

Additional options include:

# Process prompts and create a backup of the raw file
npm run prompt:process:backup

# Process prompts without removing the raw file
npm run prompt:process:keep

Organizing Prompts

To organize prompts into a structured directory hierarchy based on their tags:

# Organize prompts into category directories
npm run prompt:organize

# Preview what would be organized without making changes
npm run prompt:organize:dry

# Force reorganization (overwrite existing files)
npm run prompt:organize:force

This organizes prompts into categories like:

• development/ - Programming, coding, debugging prompts
• analysis/ - Data analysis, research, and insights prompts
• content/ - Translation and language prompts
• planning/ - Future planning, decision-making prompts
• productivity/ - Workflow and organization prompts
• ai/ - General AI and language model prompts
• templates/ - Reusable prompt templates

Managing Tags

You can manage tags across all prompts with the tag management tool:

# List all unique tags and their usage statistics
npm run prompt:tags:list

# Add a tag to all prompts matching a search term
npm run prompt:tags:add ai-assistant "You are a"

# Remove a tag from all prompts
npm run prompt:tags:remove outdated-tag

# Rename a tag across all prompts
npm run prompt:tags:rename old-tag new-tag

Exporting Prompts

You can export your prompts for sharing or backup in different formats:

# Export all prompts to a JSON file
npm run prompt:export

# Export as a ZIP archive with organized folder structure
npm run prompt:export:zip

# Export as a Markdown documentation file
npm run prompt:export:md

Additional options:

# Export only prompts with specific tags
npm run prompt:export -- --tags=ai,coding

# Specify a custom output filename
npm run prompt:export -- --out=my-collection.zip

All exports are saved to the exports/ directory with a timestamp in the filename.

Importing Prompts

You can import prompts from various sources:

# Import from a previously exported JSON file
npm run prompt:import -- --source=exports/mcp-prompts-export.json

# Import from a directory containing JSON prompt files
npm run prompt:import -- --source=path/to/prompts/

# Import from a ZIP archive
npm run prompt:import -- --source=path/to/prompts.zip

# Preview what would be imported without making changes
npm run prompt:import:dry -- --source=path/to/prompts.zip

# Force import (overwrite existing prompts)
npm run prompt:import:force -- --source=path/to/prompts.zip

Imported prompts will automatically be organized into appropriate category directories based on their tags.

Testing the Server

The project includes a test client that can verify server functionality:

npm test

This will connect to the MCP server and test all its functionality.

Claude Desktop Integration

To integrate with Claude Desktop, add the following to your Claude Desktop configuration file:

{
  "mcpServers": {
    "prompt-manager": {
      "command": "node",
      "args": ["/absolute/path/to/mcp-prompts/build/index.js"]
    }
  }
}

Available Tools

The server provides the following MCP tools:

• add_prompt: Add a new prompt to the collection
• edit_prompt: Edit an existing prompt
• get_prompt: Retrieve a prompt by ID
• list_prompts: List all prompts, optionally filtered by tags
• apply_template: Apply a template prompt with variable substitution
• delete_prompt: Delete a prompt from the collection

Using Prompts in Claude

Prompts can be accessed through Claude using MCP tools or the standard MCP prompts protocol:

I need to review some code.

use_mcp_tool({
  server_name: "prompt-manager",
  tool_name: "apply_template",
  arguments: {
    id: "code-review",
    variables: {
      language: "python",
      code: "def example():\n    return 'Hello, World!'"
    }
  }
});

Development

DevContainer Support

The project includes VS Code DevContainer configuration for consistent development environments. To use it:

1. Open the project in VS Code
2. When prompted, click "Reopen in Container"
3. VS Code will build the container and set up the environment

Project Structure

The project has been consolidated to reduce file count and improve maintainability:

• src/core/ - Core functionality consolidated into modular components

  • index.ts - Central type definitions and core functions
  • prompt-management.ts - Unified prompt management functionality
  • test.ts - Test utilities

• bin/ - Command-line tools

  • prompt-cli.js/ts - Unified CLI for all prompt management tasks
  • cli.js - Main MCP server CLI

Prompt Management

Prompts can be managed using the consolidated prompt-cli tool:

# Process raw prompts
npm run prompt:process

# Import prompts
npm run prompt:import -- --source=path/to/prompts

# Export prompts
npm run prompt:export -- --format=markdown

# Manage tags
npm run prompt:tags:list
npm run prompt:tags:add -- --tag=example --prompts=id1,id2

PGAI Integration

The MCP Prompts Server now supports PostgreSQL-based storage using PGAI (PostgreSQL AI) for enhanced prompt management capabilities.

Setting Up PostgreSQL with PGAI

For detailed setup instructions, see our PGAI Setup Guide.

Quick setup:

# Install PostgreSQL
sudo apt-get install postgresql postgresql-contrib
   
# Install vector extension for embeddings
psql -U postgres -c 'CREATE EXTENSION vector;'
   
# Install PGAI extension
psql -U postgres -c 'CREATE EXTENSION pgai;'

# Create a database for MCP Prompts
createdb -U postgres mcp_prompts

# Install the required Node.js dependencies
npm run install:deps

Configuring PGAI Storage

To use PGAI for prompt storage, update your configuration in config/pgai.json:

{
  "server": {
    "port": 3000,
    "host": "localhost",
    "logLevel": "info"
  },
  "storage": {
    "type": "pgai",
    "options": {
      "connectionString": "postgresql://username:password@localhost:5432/mcp_prompts"
    }
  }
}

Migrating Prompts to PGAI

The system provides tools to migrate prompts from file storage to the PGAI database:

# Preview migration without making changes
npm run pgai:migrate:dry

# Migrate selected prompts to PGAI
npm run pgai:migrate

# Migrate with a custom connection string
npm run pgai:migrate -- --connection=postgresql://username:password@localhost:5432/mcp_prompts

Improved Prompts Collection

The system now includes a collection of high-quality, professionally crafted prompts designed for various specialized use cases:

# Migrate the improved prompts collection to PGAI
npm run pgai:migrate:improved

# Preview the improved prompts without migrating
npm run pgai:migrate:improved:dry

The improved prompts collection includes:

Development Category

• Enhanced Code Review Assistant - Comprehensive code review with security focus
• Advanced Code Refactoring Assistant - Structured approach to code improvement
• Intelligent Debugging Assistant - Systematic problem diagnosis and resolution

Analysis Category

• Comprehensive Data Analysis Assistant - Advanced analytics for complex datasets
• Advanced Content Analysis Assistant - Sophisticated content structure analysis

Design Category

• System Architecture Designer - Professional software architecture planning

Research Category

• Comprehensive Research Assistant - Research methodology and synthesis
• Topic Modeling Specialist - Hierarchical theme identification and organization

Language Category

• Contextual Translation Assistant - Context-aware translation between languages

Planning Category

• Strategic Foresight Planner - Decision analysis and scenario planning

Productivity Category

• Question Generation Specialist - Creating diverse, thought-provoking questions
• Follow-up Question Generator - Targeted questions for deeper conversations

Testing PGAI Functionality

A test script is provided to verify the PGAI integration:

# Run PGAI integration tests
npm run pgai:test

# Test with a custom connection string
npm run pgai:test -- --connection=postgresql://username:password@localhost:5432/mcp_prompts

Benefits of PGAI Integration

Using PGAI for prompt storage provides several advantages:

1. Semantic Search: Find prompts by meaning rather than just keywords
2. Scalability: Support for larger prompt collections
3. Database Features: Transactions, concurrent access, and data integrity
4. AI Capabilities: Leverage PostgreSQL AI features for advanced prompt management

License

MIT