This guide explains how to set up and use UniAuto MCP Server with Smithery.ai to integrate with Claude, VSCode, Cursor, and other AI assistants.
Smithery.ai is a platform that connects AI assistants to tools and services, acting as a bridge between large language models and real-world applications. It enables AI models to execute actions through a standardized protocol.
npm install -g @smithery/cli
Clone the repository and install dependencies:
git clone https://github.com/yourusername/uniauto-mcp-server.git
cd uniauto-mcp-server
npm install
Alternatively, you can install the package globally:
npm install -g uniauto-mcp-server
Copy the example environment file and add your API keys:
cp .env.example .env
Edit the .env file to add your Claude API key:
# Claude API key for AI integration
CLAUDE_API_KEY=your_claude_api_key_here
CLAUDE_MODEL=claude-3-7-sonnet-20240229
# Server configuration
PORT=3000
# Start the server
npm run dev
The server will start on http://localhost:3000 (or the port you configured in .env).
# Connect UniAuto to Smithery
smithery connect uniauto-mcp-server
This will register the UniAuto MCP server with Smithery.ai.
# Connect to Claude
smithery connect --assistant claude
After connecting via Smithery, simply use Claude Web or Claude Code as usual. You can now ask Claude to:
Claude will use the UniAuto MCP server capabilities through Smithery’s connection.
Example prompt:
I'd like you to generate comprehensive Playwright tests for a login form at https://example.com/login. Make sure to test successful login, validation errors, and forgot password flow. Then run accessibility tests on the same page.
Configure the extension to use Smithery tools:
Open VSCode settings.json and add:
{
"claude.smithery.enabled": true,
"claude.smithery.path": "/path/to/smithery/executable", // Optional - only if not in PATH
"claude.smithery.tools": ["uniauto-mcp-server"]
}
Example prompt:
Can you generate a test suite for my React application? I need tests for the login component, user profile, and settings page. Use the UniAuto MCP server to analyze the application and create appropriate tests.
Enable Model Context Protocol in Cursor settings:
Open Cursor settings and add:
{
"ai.mcp.enabled": true,
"ai.mcp.tools": [
{
"name": "uniauto-mcp-server",
"url": "http://localhost:3000/api"
}
]
}
Example prompt:
I'd like to test my e-commerce application's checkout flow. Can you generate E2E tests and run performance tests on each step of the flow?
If using GitHub Copilot Chat with Claude:
Configure the Copilot extension in VSCode:
{
"github.copilot.chat.tools": [
{
"name": "UniAuto MCP",
"description": "Universal test automation with self-healing capabilities",
"endpoint": "http://localhost:3000/api/mcp/invoke",
"manifest": "http://localhost:3000/api/mcp/manifest"
}
]
}
When using UniAuto MCP server through Smithery, the AI assistants can leverage the following capabilities:
Here are the key MCP actions available to AI assistants:
Basic Commands:
- navigate: Open a URL
- click: Click on an element
- type: Enter text in a field
- extract: Get data from a page
Test Generation:
- generate_tests: Create tests for a specific framework
- generate_test_suite: Create a full suite of tests
- scaffold_project: Set up a test project structure
Advanced Testing:
- visual_compare: Compare against baseline images
- accessibility_test: Check for accessibility issues
- performance_test: Measure performance metrics
- network_trace: Monitor API and network activity
- run_test_suite: Execute a comprehensive test suite
For complete details, see the MCP manifest.
smithery list connectionssmithery list toolslogs directorysmithery logs to view Smithery connection logsDEBUG=uniauto:*If you encounter issues, please: