prompt -app

This MCP server provides tooling to manage AI prompts as MCP tools. It enables converting one-shot prompts into reusable tools, testing those tools through a clean Chat interface, and exporting them in MCP-compatible formats for integration with clients like Claude desktop, Cursor, or Cline. The server is designed to work with the T3 stack-based frontend and Prisma-backed PostgreSQL database, allowing you to create, test, and organize prompt-derived tools in projects for easy MCP client integration. Use the server to generate a prompt-to-tool workflow, validate behavior in a test chat, and export the resulting tools as MCP-compatible artifacts for downstream clients.

Prerequisites:

• Node.js >= 18.x
• pnpm >= 8.x
• Docker Desktop (optional, for local Postgres)

1. Clone the repository or download the project files.
2. Install dependencies:

pnpm install

3. Create a .env file from the example and configure environment variables (database, auth, etc.):

cp .env.example .env

4. If using a local Postgres database, start it (via Docker or your preferred method) and ensure the DATABASE_URL in .env points to it.
5. Start the development server:

pnpm run dev

6. Open the app in your browser at http://localhost:3000 to begin creating prompts, converting them to tools, and testing them.

Tips and caveats:

• Ensure the DATABASE_URL in your environment is reachable; migrations may run automatically during setup.
• The app supports exporting tools in MCP format for various clients; use the export feature to generate client-ready artifacts.
• If you modify environment variables, restart the server for changes to take effect.
• For testing prompts, use the built-in Chat interface to iterate on prompt behavior before exporting.
• If you encounter authentication issues, verify NEXTAUTH_SECRET and NEXTAUTH_URL match your deployment environment.
• When running locally, ensure port 3000 is available or adjust the NEXTAUTH_URL accordingly.