PromptArchitect

You can refine and improve your prompts with a dedicated MCP server that leverages PromptArchitect’s AI-powered prompt engineering. It tailors refinements to your workspace context, including your tech stack, project structure, and dependencies, so you get prompts that fit your codebase and goals without needing your own API key.

How to use

Set up MCP clients in your preferred editor or application that supports the Model Context Protocol. The server runs locally via a simple command and exposes a consistent interface you can invoke from your IDE, editor, or tooling. Use it to refine prompts, analyze prompt quality, and generate prompts tailored to your project.

How to install

Prerequisites: Node.js installed on your system. You will also need npm to install packages.

# Install the MCP server locally
npm install -g @merabylabs/promptarchitect-mcp

# Or install locally in your project
npm install @merabylabs/promptarchitect-mcp

Configuration and usage notes

Configure each client to point at the MCP server using the standard MCP setup. For local use, the recommended runtime command is always the same: run the MCP server via an executable command that invokes the package with its MCP interface.

{
  "mcpServers": {
    "claude_mcp": {
      "command": "npx",
      "args": ["@merabylabs/promptarchitect-mcp"]
    },
    "vscode Copilot_mcp": {
      "command": "npx",
      "args": ["@merabylabs/promptarchitect-mcp"]
    }
  }
}

Programmatic usage

Access the server’s capabilities from code to refine prompts, analyze prompts, and generate prompts programmatically.

import { refinePrompt, analyzePrompt } from '@promptarchitect/mcp-server';

const result = await refinePrompt({
  prompt: 'Write code to sort an array',
  feedback: 'Make it more specific about language and edge cases',
});

console.log(result.refinedPrompt);

const analysis = await analyzePrompt({
  prompt: 'Help me with my code',
});
console.log(analysis.scores);
console.log(analysis.suggestions);

Examples of supported clients and how to pair them

The MCP server works with a variety of editors and IDEs. You configure the MCP server once per client and then use the client’s prompt refinement features as part of your normal workflow.

Troubleshooting

If you encounter connection issues, verify that the MCP command is accessible from your shell and that the editor’s MCP integration is configured to point to the same command and arguments used above.

Available tools

refinePrompt

Improve your current prompt based on feedback and workspace context

analyzePrompt

Evaluate prompt quality with scores and improvement suggestions

generatePrompt

Transform a raw idea into a well-structured prompt tailored to your project