Skip to main content
Glama

MCP Server for Crawl4AI

by omgwtfwow
capture-screenshot.integration.test.ts4.18 kB
/* eslint-env jest */ import { Client } from '@modelcontextprotocol/sdk/client/index.js'; import { createTestClient, cleanupTestClient, TEST_TIMEOUTS } from './test-utils.js'; interface ToolResult { content: Array<{ type: string; text?: string; data?: string; mimeType?: string; }>; } describe('capture_screenshot Integration Tests', () => { let client: Client; beforeAll(async () => { client = await createTestClient(); }, TEST_TIMEOUTS.medium); afterAll(async () => { if (client) { await cleanupTestClient(client); } }); describe('Screenshot capture', () => { it( 'should capture screenshot with default wait time', async () => { const result = await client.callTool({ name: 'capture_screenshot', arguments: { url: 'https://httpbin.org/html', }, }); expect(result).toBeDefined(); const content = (result as ToolResult).content; expect(content).toHaveLength(2); // First item should be the image expect(content[0].type).toBe('image'); expect(content[0].mimeType).toBe('image/png'); expect(content[0].data).toBeTruthy(); expect(content[0].data?.length).toBeGreaterThan(1000); // Should be a substantial base64 string // Second item should be text description expect(content[1].type).toBe('text'); expect(content[1].text).toContain('Screenshot captured for: https://httpbin.org/html'); }, TEST_TIMEOUTS.short, ); it( 'should capture screenshot with custom wait time', async () => { const result = await client.callTool({ name: 'capture_screenshot', arguments: { url: 'https://httpbin.org/html', screenshot_wait_for: 0.5, // Reduced from 3 seconds }, }); expect(result).toBeDefined(); const content = (result as ToolResult).content; expect(content).toHaveLength(2); // First item should be the image expect(content[0].type).toBe('image'); expect(content[0].mimeType).toBe('image/png'); expect(content[0].data).toBeTruthy(); // Second item should be text description expect(content[1].type).toBe('text'); expect(content[1].text).toContain('Screenshot captured for: https://httpbin.org/html'); }, TEST_TIMEOUTS.medium, ); it( 'should reject session_id parameter', async () => { const result = await client.callTool({ name: 'capture_screenshot', arguments: { url: 'https://example.com', session_id: 'test-session', }, }); const content = (result as ToolResult).content; expect(content).toHaveLength(1); expect(content[0].type).toBe('text'); expect(content[0].text).toContain('session_id'); expect(content[0].text).toContain('does not support'); expect(content[0].text).toContain('stateless'); }, TEST_TIMEOUTS.short, ); it( 'should handle invalid URLs gracefully', async () => { const result = await client.callTool({ name: 'capture_screenshot', arguments: { url: 'not-a-valid-url', }, }); const content = (result as ToolResult).content; expect(content).toHaveLength(1); expect(content[0].type).toBe('text'); expect(content[0].text).toContain('Error'); expect(content[0].text?.toLowerCase()).toContain('invalid'); }, TEST_TIMEOUTS.short, ); it( 'should handle non-existent domains', async () => { const result = await client.callTool({ name: 'capture_screenshot', arguments: { url: 'https://this-domain-definitely-does-not-exist-123456789.com', }, }); const content = (result as ToolResult).content; expect(content).toHaveLength(1); expect(content[0].type).toBe('text'); expect(content[0].text).toContain('Error'); }, TEST_TIMEOUTS.short, ); }); });

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/omgwtfwow/mcp-crawl4ai-ts'

If you have feedback or need assistance with the MCP directory API, please join our Discord server