Prompt Manager

May 23, 2025 ยท View on GitHub

A standalone system for managing AI prompts with composable elements.

Overview

Prompt Manager provides a web interface and API for creating, organizing, and managing prompts for AI interactions. It supports composable templates, prompt directories, and real-time editing. There is a focus on keyboard shortcuts and efficient workflow- directory listing and prompt embedding support tab completion, most interactions can be navigated using keyboard shortcuts, etc.

Features

  • Create and manage prompts (standard and composite)
  • Compose templates with prompt inclusions and expansions
  • Real-time editing via WebSocket (for prompts)
  • Access prompts via RESTful API
  • Render templates with all inclusions expanded
  • Organize prompts in directories
  • Integrates with Model Context Protocol (MCP) for advanced prompt workflows

Quick Start

  1. Clone the repository:

    git clone <repository-url>
cd prompt_manager

  2. Set up the environment (choose one):

    # Using Taskfile (recommended for full dev setup)
task setup:py
task setup:js
task setup:e2e
# Or using Makefile
make setup

  3. Run the application:

    ./bin/start_prompt_manager.sh
# Or
bin/restart_prompt_manager.sh

  4. Open your browser to http://localhost:8081

Usage

The prompt manager can be used in two ways:

  1. Web Interface: Access the web UI at http://localhost:8081
  2. API: Use the RESTful API endpoints to programmatically manage prompts

API Endpoints

Prompt Management

  • GET /api/prompts/all: List all prompts
  • GET /api/prompts/{prompt_id}: Get a specific prompt
  • POST /api/prompts: Create a new prompt
  • PUT /api/prompts/{prompt_id}: Update a prompt
  • DELETE /api/prompts/{prompt_id}: Delete a prompt
  • POST /api/prompts/expand: Render a template with inclusions expanded
  • POST /api/prompts/rename: Rename a prompt

Directory Management

  • GET /api/prompts/directories/all: List all prompt directories
  • POST /api/prompts/directories: Add a new directory
  • DELETE /api/prompts/directories/{directory_path}: Remove a directory
  • POST /api/prompts/reload: Reload all prompts from all directories

WebSocket Interface

  • Connect to /api/prompts/ws/{prompt_id} for real-time prompt editing
  • Uses standard WebSocket protocol with JSON messages for updates

Configuration

The system uses several directories for prompt storage:

  • Project data directory: prompt_manager/data/prompts/
  • User config directory: ~/.prompt_manager/prompts/

MCP Integration

Prompt Manager supports integration with the Model Context Protocol (MCP) for advanced prompt workflows and Claude Desktop compatibility. See mcp_server/ and related scripts for details.

Development & Testing

  • Run tests with pytest or using the provided Taskfile/Makefile targets.
  • See CHANGES.md for recent updates and protocol fixes.
  • See README.md in mcp_server/ for MCP-specific details.

License

Apache License 2.0. See LICENSE.