MCP Test Kit

A comprehensive testing framework for Model Context Protocol (MCP) servers

📖 Table of Contents

• What is MCP?
• What is MCP Test Kit?
• Why Use MCP Test Kit?
• Installation
• Quick Start
• Core Concepts
• API Reference
• Examples
• Best Practices
• Contributing
• License

🤔 What is MCP?

Model Context Protocol (MCP) is an open protocol developed by Anthropic that enables AI assistants like Claude to securely interact with external tools, data sources, and services. Think of it as a standardized way for AI models to:

• Access tools: Execute functions like searching databases, calling APIs, or running calculations
• Read resources: Fetch data from files, databases, or external services
• Use prompts: Leverage pre-built prompt templates for common tasks

Real-World MCP Examples

Here are some practical MCP servers you might build:

1. Weather Server: Provides current weather data for any city
2. Database Server: Allows querying a PostgreSQL/MySQL database
3. File System Server: Enables reading/writing files on disk
4. API Integration Server: Connects to third-party APIs (GitHub, Slack, etc.)
5. Calculator Server: Performs complex mathematical operations

How MCP Works

graph LR
    A[Claude<br/>Client] <-->|MCP| B[MCP Server<br/>Your Code]
    B <--> C[Your Data<br/>or Service]

The MCP server acts as a bridge between Claude and your data/services, exposing them through a standardized protocol.

🎯 What is MCP Test Kit?

MCP Test Kit is a testing framework that makes it easy to write automated tests for your MCP servers. It provides:

🔧 Core Features

1. MCPTestClient: A test client that connects to your MCP server and calls its tools/resources
2. Custom Vitest Matchers: Specialized assertions for validating MCP responses
3. Mock Server: Create fake MCP servers for testing client code
4. TypeScript Support: Full type definitions for better IDE support and type safety

🎪 The Problem It Solves

Without MCP Test Kit, testing an MCP server is challenging:

• ❌ You need to manually start the server and interact with it
• ❌ No standardized way to assert MCP-specific behaviors
• ❌ Complex setup for integration tests
• ❌ Hard to test error conditions and edge cases

With MCP Test Kit:

• ✅ Programmatically start and stop your server in tests
• ✅ Use intuitive matchers like toBeValidMCPTool() and toMatchMCPToolResponse()
• ✅ Write comprehensive integration tests in minutes
• ✅ Easily test error handling and edge cases

💡 Why Use MCP Test Kit?

For MCP Server Developers

If you're building an MCP server, MCP Test Kit helps you:

• Verify correctness: Ensure your server implements the MCP protocol correctly
• Test tools: Validate that your tools return the expected results
• Test resources: Confirm resources are accessible and return proper data
• Test error handling: Verify error responses follow MCP standards
• Regression testing: Catch bugs before they reach production
• Documentation: Tests serve as living documentation of your server's behavior

Example Use Case

Let's say you built a weather MCP server. With MCP Test Kit, you can write tests like:

it('should return weather data for San Francisco', async () => {
  const result = await client.callTool('get-weather', { 
    city: 'San Francisco' 
  });
  
  expect(result).toMatchMCPToolResponse();
  expect(result.content[0].text).toContain('temperature');
  expect(result.content[0].text).toContain('San Francisco');
});

it('should handle invalid city names', async () => {
  await expect(
    client.callTool('get-weather', { city: '' })
  ).rejects.toMatchMCPError({
    code: -32602,
    message: /invalid.*city/i
  });
});

📦 Installation

# Using npm
npm install --save-dev @crashbytes/mcp-test-kit

# Using yarn
yarn add -D @crashbytes/mcp-test-kit

# Using pnpm
pnpm add -D @crashbytes/mcp-test-kit

Prerequisites

• Node.js 18+
• A testing framework (Vitest recommended)
• TypeScript (optional, but recommended)

🚀 Quick Start

1. Create Your First Test

// tests/weather-server.test.ts
import { describe, it, expect, beforeAll, afterAll } from 'vitest';
import { createMCPTestClient } from '@crashbytes/mcp-test-kit';
import type { MCPTestClient } from '@crashbytes/mcp-test-kit';
import '@crashbytes/mcp-test-kit/matchers';

describe('Weather MCP Server', () => {
  let client: MCPTestClient;

  beforeAll(async () => {
    // Connect to your MCP server
    client = await createMCPTestClient({
      command: 'node',
      args: ['dist/server.js'],
    });
  });

  afterAll(async () => {
    await client.disconnect();
  });

  it('should list available tools', async () => {
    const tools = await client.listTools();
    
    expect(tools).toHaveLength(1);
    expect(tools[0]).toBeValidMCPTool();
    expect(tools[0].name).toBe('get-weather');
  });

  it('should get weather for a city', async () => {
    const result = await client.callTool('get-weather', {
      city: 'London',
    });

    expect(result).toMatchMCPToolResponse();
    expect(result.content[0].text).toContain('London');
  });
});

2. Run Your Tests

npm test

That's it! You're now testing your MCP server.

🧩 Core Concepts

MCPTestClient

The MCPTestClient is your main interface for testing MCP servers. It:

• Spawns your MCP server as a subprocess
• Establishes a connection using the MCP protocol
• Provides methods to call tools, list resources, etc.
• Handles cleanup when tests complete

Key Methods:

• listTools(): Get all available tools from the server
• callTool(name, args): Execute a tool with arguments
• listResources(): Get all available resources
• readResource(uri): Read a specific resource
• listPrompts(): Get all available prompts
• getPrompt(name, args): Get a specific prompt

Custom Matchers

MCP Test Kit provides custom Vitest matchers for MCP-specific assertions:

toBeValidMCPTool()

Validates that an object is a properly formatted MCP tool:

const tool = {
  name: 'calculate',
  description: 'Performs calculations',
  inputSchema: {
    type: 'object',
    properties: {
      expression: { type: 'string' }
    }
  }
};

expect(tool).toBeValidMCPTool();

toBeValidMCPResource()

Validates that an object is a properly formatted MCP resource:

const resource = {
  uri: 'file:///data/users.json',
  name: 'Users Database',
  description: 'List of all users'
};

expect(resource).toBeValidMCPResource();

toMatchMCPToolResponse()

Validates the structure of a tool response:

const response = await client.callTool('get-weather', { city: 'Paris' });

expect(response).toMatchMCPToolResponse();
expect(response.content[0].text).toContain('Paris');

toMatchMCPError(error)

Validates MCP error responses:

await expect(
  client.callTool('invalid-tool', {})
).rejects.toMatchMCPError({
  code: -32601, // Method not found
  message: /not found/i
});

toHaveMCPProtocolVersion(version)

Validates the MCP protocol version:

const info = await client.getServerInfo();
expect(info).toHaveMCPProtocolVersion('2024-11-05');

📚 API Reference

createMCPTestClient(config)

Creates and connects to an MCP test client.

Parameters:

interface MCPTestClientConfig {
  command: string;           // Command to run (e.g., 'node', 'python')
  args?: string[];          // Arguments for the command
  env?: Record<string, string>; // Environment variables
  timeout?: number;         // Timeout in milliseconds (default: 5000)
  transport?: 'stdio';      // Transport type (only stdio supported)
  debug?: boolean;          // Enable debug logging
}

Returns: Promise<MCPTestClient>

Example:

const client = await createMCPTestClient({
  command: 'node',
  args: ['dist/server.js'],
  env: {
    NODE_ENV: 'test',
    DATABASE_URL: 'sqlite::memory:'
  },
  timeout: 10000,
  debug: true
});

MCPTestClient Methods

listTools(): Promise<MCPTool[]>

Lists all tools available on the MCP server.

const tools = await client.listTools();
console.log(tools[0].name); // 'get-weather'

callTool(name: string, args?: object): Promise<MCPToolResult>

Executes a tool with the given arguments.

const result = await client.callTool('calculate', {
  expression: '2 + 2'
});

listResources(): Promise<MCPResource[]>

Lists all resources available on the MCP server.

const resources = await client.listResources();
console.log(resources[0].uri); // 'file:///data/users.json'

readResource(uri: string): Promise<MCPResourceContent>

Reads the content of a specific resource.

const content = await client.readResource('file:///data/users.json');
console.log(content.text);

listPrompts(): Promise<MCPPrompt[]>

Lists all prompts available on the MCP server.

const prompts = await client.listPrompts();

getPrompt(name: string, args?: Record<string, string>): Promise<unknown>

Gets a specific prompt with arguments.

const prompt = await client.getPrompt('code-review', {
  language: 'typescript'
});

disconnect(): Promise<void>

Disconnects from the MCP server and cleans up resources.

await client.disconnect();

💻 Examples

Example 1: Testing a Calculator Server

import { describe, it, expect, beforeAll, afterAll } from 'vitest';
import { createMCPTestClient, MCPTestClient } from '@crashbytes/mcp-test-kit';
import '@crashbytes/mcp-test-kit/matchers';

describe('Calculator MCP Server', () => {
  let client: MCPTestClient;

  beforeAll(async () => {
    client = await createMCPTestClient({
      command: 'node',
      args: ['dist/calculator-server.js'],
    });
  });

  afterAll(async () => {
    await client.disconnect();
  });

  describe('Basic Operations', () => {
    it('should add two numbers', async () => {
      const result = await client.callTool('calculate', {
        operation: 'add',
        a: 5,
        b: 3
      });

      expect(result).toMatchMCPToolResponse();
      expect(result.content[0].text).toBe('8');
    });

    it('should handle division by zero', async () => {
      await expect(
        client.callTool('calculate', {
          operation: 'divide',
          a: 10,
          b: 0
        })
      ).rejects.toMatchMCPError({
        code: -32602,
        message: /division by zero/i
      });
    });
  });

  describe('Tool Validation', () => {
    it('should have valid tool schema', async () => {
      const tools = await client.listTools();
      const calcTool = tools.find(t => t.name === 'calculate');

      expect(calcTool).toBeValidMCPTool();
      expect(calcTool?.inputSchema.properties).toHaveProperty('operation');
      expect(calcTool?.inputSchema.properties).toHaveProperty('a');
      expect(calcTool?.inputSchema.properties).toHaveProperty('b');
    });
  });
});

Example 2: Testing a Database Server

import { describe, it, expect, beforeAll, afterAll } from 'vitest';
import { createMCPTestClient, MCPTestClient } from '@crashbytes/mcp-test-kit';
import '@crashbytes/mcp-test-kit/matchers';

describe('Database MCP Server', () => {
  let client: MCPTestClient;

  beforeAll(async () => {
    client = await createMCPTestClient({
      command: 'node',
      args: ['dist/db-server.js'],
      env: {
        DATABASE_URL: 'sqlite::memory:',
        NODE_ENV: 'test'
      }
    });
  });

  afterAll(async () => {
    await client.disconnect();
  });

  it('should query users from database', async () => {
    const result = await client.callTool('query', {
      sql: 'SELECT * FROM users WHERE age > 18'
    });

    expect(result).toMatchMCPToolResponse();
    const data = JSON.parse(result.content[0].text);
    expect(Array.isArray(data)).toBe(true);
  });

  it('should list database resources', async () => {
    const resources = await client.listResources();

    expect(resources.length).toBeGreaterThan(0);
    resources.forEach(resource => {
      expect(resource).toBeValidMCPResource();
    });
  });

  it('should read table schema', async () => {
    const content = await client.readResource('schema://users');
    
    expect(content.text).toContain('id');
    expect(content.text).toContain('name');
    expect(content.text).toContain('email');
  });
});

Example 3: Testing with Mocks

import { describe, it, expect } from 'vitest';
import { createMockMCPServer, mockTool } from '@crashbytes/mcp-test-kit/mocks';

describe('Mock MCP Server', () => {
  it('should create a mock server with tools', async () => {
    const server = createMockMCPServer({
      tools: [
        mockTool(
          'greet',
          async (args) => ({
            content: [{ type: 'text', text: `Hello, ${args.name}!` }]
          }),
          {
            description: 'Greets a person',
            inputSchema: {
              type: 'object',
              properties: {
                name: { type: 'string' }
              }
            }
          }
        )
      ]
    });

    const tools = server.getTools();
    expect(tools).toHaveLength(1);
    expect(tools[0].name).toBe('greet');

    const result = await server.callTool('greet', { name: 'World' });
    expect(result.content[0].text).toBe('Hello, World!');
  });
});

🎓 Best Practices

1. Use beforeAll/afterAll for Connection Management

let client: MCPTestClient;

beforeAll(async () => {
  client = await createMCPTestClient({ /* config */ });
});

afterAll(async () => {
  await client.disconnect();
});

2. Test Both Success and Failure Cases

// Test success
it('should return data for valid input', async () => {
  const result = await client.callTool('tool', { valid: true });
  expect(result).toMatchMCPToolResponse();
});

// Test failure
it('should reject invalid input', async () => {
  await expect(
    client.callTool('tool', { invalid: true })
  ).rejects.toMatchMCPError({ code: -32602 });
});

3. Validate Tool Schemas

it('should have properly defined tools', async () => {
  const tools = await client.listTools();
  
  tools.forEach(tool => {
    expect(tool).toBeValidMCPTool();
    expect(tool.description).toBeTruthy();
    expect(tool.inputSchema.properties).toBeDefined();
  });
});

4. Use Descriptive Test Names

// ❌ Bad
it('test 1', async () => { /* ... */ });

// ✅ Good
it('should return weather data for valid city names', async () => { /* ... */ });

5. Isolate Tests

Each test should be independent and not rely on state from other tests.

// ❌ Bad - Tests depend on order
it('creates a user', async () => { /* ... */ });
it('updates the user', async () => { /* assumes user exists */ });

// ✅ Good - Each test is independent
beforeEach(async () => {
  await client.callTool('reset-database', {});
});

it('creates a user', async () => { /* ... */ });
it('updates a user', async () => {
  await client.callTool('create-user', { name: 'Test' });
  await client.callTool('update-user', { name: 'Updated' });
});

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Development Setup

# Clone the repository
git clone https://github.com/crashbytes/mcp-test-kit.git
cd mcp-test-kit

# Install dependencies
npm install

# Build the project
npm run build

# Run tests
npm test

# Run linter
npm run lint

📄 License

MIT © Blackhole Software LLC

🔗 Links

• NPM Package: https://www.npmjs.com/package/@crashbytes/mcp-test-kit
• GitHub: https://github.com/crashbytes/mcp-test-kit
• MCP Specification: https://spec.modelcontextprotocol.io/
• CrashBytes Blog: https://crashbytes.com

🙏 Acknowledgments

• Anthropic for creating the Model Context Protocol
• Vitest for the excellent testing framework
• The MCP community for feedback and contributions

---

Built with ❤️ by CrashBytes