agentleFS
Sign inStart free
LucasCooperBeyLucasCooperBey/claude-skills/skills/mcp-builder/SKILL.md

mcp-builder

Skill1 starsChanged 5 months agoMITView on GitHub

Builds MCP (Model Context Protocol) servers for API integration. Guides the full development lifecycle from research to evaluation.

  • Reads credentials

We split this skill into the rules it gives an agent

5 rules. 0 are common, 5 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 174 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Test with MCP Inspector

    line 24
  2. Distinctive:

    Create 10 complex, realistic test questions

    line 28
  3. Distinctive:

    Verify LLMs can effectively use your server

    line 29
  4. Distinctive:

    Before implementing, use the package-search MCP to research SDK patterns:

    line 84
  5. Distinctive:

    Read-only (doesn't modify external state)

    line 140

“Common” is measured: kept by at least 6 in 10 of the 174 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

MCP Server Development

Build Model Context Protocol servers that enable language models to interact with external services.

Four-Phase Approach

Phase 1: Deep Research & Planning

  • Study MCP design principles and protocol documentation
  • Select tech stack (TypeScript recommended)
  • Analyze target API thoroughly
  • Plan tool organization

Phase 2: Implementation

  • Project setup with proper dependencies
  • Core infrastructure (auth, error handling)
  • Individual tools with schemas and descriptions

Phase 3: Review & Testing

  • Code quality verification
  • Test with MCP Inspector
  • Error handling validation

Phase 4: Evaluation

  • Create 10 complex, realistic test questions
  • Verify LLMs can effectively use your server

Project Setup (TypeScript)

mkdir my-mcp-server && cd my-mcp-server
bun init
bun add @modelcontextprotocol/sdk zod
// src/index.ts
import { Server } from '@modelcontextprotocol/sdk/server/index.js';
import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js';

const server = new Server(
  { name: 'my-server', version: '1.0.0' },
  { capabilities: { tools: {} } }
);

// Define tools
server.setRequestHandler('tools/list', async () => ({
  tools: [
    {
      name: 'my_tool',
      description: 'Does something useful',
      inputSchema: {
        type: 'object',
        properties: {
          param: { type: 'string', description: 'Input parameter' }
        },
        required: ['param']
      }
    }
  ]
}));

server.setRequestHandler('tools/call', async (request) => {
  const { name, arguments: args } = request.params;

  if (name === 'my_tool') {
    // Implement tool logic
    return { content: [{ type: 'text', text: 'Result' }] };
  }

  throw new Error(`Unknown tool: ${name}`);
});

// Start server
const transport = new StdioServerTransport();
await server.connect(transport);

Research with Package Search MCP

Before implementing, use the package-search MCP to research SDK patterns:

# Find how the official MCP SDK structures servers
package_search_hybrid("@modelcontextprotocol/sdk server setup patterns", registry="npm")

# Look at real tool implementations
package_search_grep("setRequestHandler.*tools/call", package="@modelcontextprotocol/sdk", registry="npm")

This grounds your implementation in real source code rather than hallucinated patterns.

Key Design Principles

Tool Naming: Consistent, action-oriented

  • Good: github_create_issue, slack_send_message
  • Bad: doThing, handler1

Error Messages: Provide specific suggestions

throw new Error(
  `Repository not found: ${repo}. ` +
  `Check that the repository exists and you have access. ` +
  `Try: gh repo view ${repo}`
);

Balance Coverage: Combine comprehensive API endpoint coverage with specialized workflow tools.

Input Validation with Zod

import { z } from 'zod';

const CreateIssueSchema = z.object({
  repo: z.string().describe('Repository in owner/name format'),
  title: z.string().min(1).describe('Issue title'),
  body: z.string().optional().describe('Issue body in markdown'),
  labels: z.array(z.string()).optional().describe('Labels to apply')
});

// In tool handler
const params = CreateIssueSchema.parse(args);

Testing with MCP Inspector

bun run build
npx @modelcontextprotocol/inspector dist/index.js

Evaluation Questions

Each evaluation question must be:

  • Independent (no dependencies on other questions)
  • Read-only (doesn't modify external state)
  • Complex enough to require multiple tool calls
  • Realistic (matches actual use cases)
  • Verifiable (single clear answer)
  • Temporally stable (answer doesn't change)

Example:

"Find all open issues in repo X labeled 'bug' and summarize their common themes"

Claude Code Integration

Add to ~/.claude/claude_desktop_config.json:

{
  "mcpServers": {
    "my-server": {
      "command": "node",
      "args": ["/path/to/dist/index.js"],
      "env": {
        "API_KEY": "your-key"
      }
    }
  }
}

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free