Introduction to MCP Resources and Prompts Sampling

steps

1. Initialize a TypeScript Project

   bashCopy

   mkdir mcp-prompts-example
   cd mcp-prompts-example
   npm init -y
   npm install typescript --save-dev
   npx tsc --init

2. Install MCP Dependencies

   bashCopy

   npm install @mcp/server @mcp/prompts

3. Define a Prompt Template

   Create a file, prompts.ts, with the following content:

   typescriptCopy

   import { createPrompt } from '@mcp/prompts';
   
   export const fetchGitHubComments = createPrompt({
       name: 'FetchGitHubComments',
       description: 'Fetches comments from a GitHub pull request.',
       parameters: {
           repo: { type: 'string', description: 'Repository name' },
           pullRequestId: { type: 'number', description: 'Pull request ID' },
       },
       execute: async ({ repo, pullRequestId }) => {
           // Simulate fetching comments - replace with actual API calls as needed
           return `Fetching comments for PR ${pullRequestId} in repo ${repo}...`;
       },
   });

4. Integrate the Prompt into Your Server

   Create an index.ts file and set up a simple MCP server using the prompt:

   typescriptCopy

   import { createServer } from '@mcp/server';
   import { fetchGitHubComments } from './prompts';
   
   const server = createServer({
       name: 'GitHub Comment Fetcher',
       description: 'A server that fetches GitHub pull request comments.',
       prompts: [fetchGitHubComments],
   });
   
   server.listen(3000, () => {
       console.log('MCP Server is running on http://localhost:3000');
   });

5. Test the Server

   Run the server, then use an MCP client to invoke the FetchGitHubComments prompt.

   bashCopy

   npx ts-node index.ts
   # MCP Server is running on http://localhost:3000

steps

1. Install PostgreSQL Driver

   bashCopy

   npm install pg

2. Create a Resource for Database Schema

   In a new file resources.ts, define the resource:

   typescriptCopy

   import { createResource } from '@mcp/server';
   
   export const databaseSchema = createResource({
       name: 'DatabaseSchema',
       description: 'Provides the schema of a PostgreSQL database.',
       async fetch() {
           return {
               tables: [
                   { name: 'users', columns: ['id', 'name', 'email'] },
                   { name: 'orders', columns: ['id', 'user_id', 'total'] },
               ],
           };
       },
   });

3. Register the Resource in Your Server

   Update index.ts:

   typescriptCopy

   import { createServer } from '@mcp/server';
   import { databaseSchema } from './resources';
   
   const server = createServer({
       name: 'Database Schema Viewer',
       description: 'A server that exposes database schema as a resource.',
       resources: [databaseSchema],
   });
   
   server.listen(3000, () => {
       console.log('MCP Server is running on http://localhost:3000');
   });

4. Test Resource Exposure

   Start the server and access the resource via an MCP client, adding it to the context window or using it for further processing.

   bashCopy

   npx ts-node index.ts
   # MCP Server is running on http://localhost:3000