promptcode-preset-mcp

A CLI tool for managing file presets for code prompting. Quickly assemble collections of files from your codebase into prompts for AI/LLM interactions.

Installation

Global installation (recommended)

# Install from npm
pnpm install -g @cogflows/promptcode-preset-mcp

# Or install from GitHub
pnpm install -g github:cogflows/promptcode-preset-mcp

Local development

# Clone the repository
git clone https://github.com/cogflows/promptcode-preset-mcp.git
cd promptcode-preset-mcp

# Install dependencies
pnpm install

# Build and link globally
pnpm build
pnpm link --global

Usage

List presets

promptcode ls

Get preset content

# Export to temp file and print path
promptcode get <preset-name>

# Export and open in default editor
promptcode get <preset-name> --open

Set workspace directory

# Use environment variable
WORKSPACE=/path/to/project promptcode ls
WORKSPACE=/path/to/project promptcode get core

Creating Presets

Presets are JSON files stored in .promptcode/presets/ within your workspace directory.

Example preset file .promptcode/presets/core.json:

{
  "name": "core",
  "files": [
    "src/main.ts",
    "src/utils.ts",
    "README.md"
  ]
}

Example Workflow

1. Create a presets directory in your project:

   mkdir -p .promptcode/presets
   

2. Create preset files:

   echo '{
     "name": "api",
     "files": ["src/api/index.ts", "src/api/routes.ts", "src/types.ts"]
   }' > .promptcode/presets/api.json
   

3. List available presets:

   promptcode ls
   

   Output:

   [
     {
       "name": "api",
       "fileCount": 3,
       "totalTokens": 1250
     }
   ]
   

4. Export preset content:

   promptcode get api
   

   Output: /var/folders/.../T/preset-api-2025-06-02T19-12-41-238Z.txt

5. Use the exported file with your AI tool:

   # Copy to clipboard (macOS)
   cat $(promptcode get api) | pbcopy
   
   # Or open directly
   promptcode get api --open
   

Features

• Token counting: Uses OpenAI's tiktoken to count tokens in files
• Caching: Token counts are cached for performance
• Workspace support: Work with presets from any project directory
• Clean output: Returns just the file path for easy scripting

Development

The project is written in TypeScript and uses:

• Node.js ES modules
• tiktoken for token counting
• TypeScript for type safety

Project Structure

src/
├── cli.ts           # CLI interface
├── presetManager.ts # Preset loading/saving logic
├── tokenCounter.ts  # Token counting with caching
└── types/
    └── filePreset.ts # TypeScript interfaces