mcp-prompts

MCP Prompts is a versatile MCP server that manages, versions, and serves prompts and templates for LLM workflows. It supports multiple storage backends (memory, file, AWS) and can operate in MCP stdio mode or as an HTTP REST API, making it suitable for both local development and production deployments. The server exposes a rich set of MCP tools for prompt management and template handling, including creating, reading, updating, deleting, and applying templates to prompts, along with search, tagging, and access control features. When running in stdio mode, it integrates with MCP clients via standard input/output; when running as an HTTP server, it exposes REST endpoints for programmatic access to prompts, templates, and related analytics. This bifurcation allows you to embed MCP Prompts in client apps, automation scripts, or web services while maintaining consistent MCP semantics.

Prerequisites:

• Node.js (LTS) and npm or pnpm yarn as preferred package manager
• Access to npm registry (public or private)

Install (per-project, no global install required):

1. Create a project folder and initialize: mkdir mcp-prompts-usage && cd mcp-prompts-usage npm init -y
2. Install the MCP Prompts package (as a dependency): npm install @sparesparrow/mcp-prompts
3. Use the MCP Prompts CLI or programmatic API as documented by the package.

If you prefer the CLI globally:

1. Install globally: npm install -g @sparesparrow/mcp-prompts
2. Run the CLI helper to inspect usage: mcp-prompts --help

Docker (optional):

• Pull the official image: docker pull ghcr.io/sparesparrow/mcp-prompts:latest
• Run in MCP mode (stdio): docker run -i --rm ghcr.io/sparesparrow/mcp-prompts:latest
• Run as HTTP server (example): docker run -p 3000:3000 -e MODE=http ghcr.io/sparesparrow/mcp-prompts:latest

Environment configuration is flexible: you can switch between MCP stdio and HTTP server modes using the MODE environment variable (mcp for stdio, http for REST API). When using AWS storage, ensure proper AWS credentials and region configuration. For local testing, the in-memory storage backend is recommended. If you deploy with Docker, consider mounting a persistent volume for data directories when using file storage. Review the AWS and Stripe-related environment variables if you enable AWS storage or payment processing. The MCP tools exposed by the server include add_prompt, get_prompt, list_prompts, update_prompt, delete_prompt, apply_template, and get_stats, along with a robust template system for variable substitution.