MCP Prompt Server

You run an MCP server that serves preset prompt templates as tools for editors like Cursor and Windsurf. This lets you manage, customize, and reuse prompts to automate tasks such as code review and API documentation, improving consistency and collaboration across different AI agents.

How to use

Once your MCP client connects to the Prompt Server, you can browse and invoke the available prompt tools to perform common tasks. The server exposes prompts as tools you can call with specific arguments. For example, you can request a code review or generate API documentation by selecting the corresponding prompt tool and supplying the required inputs. The prompts support dynamic parameters so you can tailor the output to your code language, project type, or preferred format. You can also reload prompts after adding new templates to keep your toolkit up to date.

How to install

1) Prerequisites
   - Install Node.js (recommended LTS) and npm on your system.
   - Ensure you have access to the machine where you will run the MCP server.

2) Install and start the Prompt Server
   - Clone or obtain the Prompt Server source in your workspace.
   - Install dependencies

cd prompt-server
npm install

3. Run the server

npm start

The server will start and listen for connections from MCP clients like Cursor and Windsurf.

## Configuration and usage notes

You can configure your editors to connect to the Prompt Server via standard MCP transport. The following examples illustrate how to wire the local server so your editors can access the prompt tools. You can place the configuration in your editor’s MCP settings as shown.

// Cursor MCP config example { "servers": [ { "name": "Prompt Server", "command": ["node", "/path/to/prompt-server/src/index.js"], "transport": "stdio", "initialization_options": {} } ] }

// Windsurf MCP config example { "mcpServers": { "prompt-server": { "command": "node", "args": [ "/path/to/prompt-server/src/index.js" ], "transport": "stdio" } } }

## Adding new prompt templates

You can extend the available prompts by adding new YAML or JSON templates in the prompts directory. Each template should define a unique name, a description, optional arguments, and the sequence of messages that compose the prompt. After adding new files, the server will load them on startup or you can trigger a prompt reload using the provided tooling.

name: prompt_name description: prompt description arguments:

• name: arg_name description: arg description required: true messages:
• role: user content: type: text text: |- Your prompt text here with {{arg_name}} placeholders

## Available tools

### code\_review

Prompts for code review tasks that analyze and suggest improvements for code snippets in multiple languages.

### api\_documentation

Templates to generate API documentation from code or schemas, outputting in a chosen format such as Markdown.

### code\_refactoring

Guided prompts that propose refactorings to improve code structure, readability, and maintainability.

### test\_case\_generator

Prompts to produce test cases based on code snippets or requirements to assist testing workflows.

### project\_architecture

Templates that outline or generate high-level project architecture descriptions and diagrams.