Selenium MCP Server with AI

This JAVA MAVEN TESTNG Project implements a Selenium Master Control Program (MCP) server in Java that executes browser automation tasks by converting natural language test steps into Selenium WebDriver commands using a Language Model (LLM). The server supports local Ollama or remote LLM endpoints and is configurable via property files. A TestNG-based client is included to send test prompts and log results.

Project Structure

• src/main/java/McpServer.java: Main server class that handles HTTP requests, LLM interaction, and Selenium command execution.
• src/main/java/LlmClient.java: Handles communication with the LLM (Ollama or remote API).
• src/test/java/SeleniumTestClient.java: TestNG client that sends test steps to the server and logs results.
• config.properties: Configuration file for server settings (port, browser, LLM endpoint).
• selenium_operations.properties: Defines supported Selenium operations and their required parameters.
• pom.xml: Maven configuration with dependencies.

Application Flow:

• Step 1: Client input natural language test steps to MCP Server
• Step 2: MCP Server processes the input and LLM is called with an augmented PROMPT
• Step 3: LLM Responds with command for MCP
• Step 4: MCP Server uses the response and perform Selenium commands on browser

Prerequisites

• Java: 17 or later
• Maven: 3.6.0 or later
• WebDriver: ChromeDriver or FirefoxDriver (matching your browser version) installed and added to PATH
• LLM: Either:
  • Local Ollama running on http://localhost:11434 (default) with a model like llama3
  • A remote LLM endpoint (e.g., OpenAI-compatible API) with an API key
• Browser: Chrome or Firefox installed
• TestNG: Included via Maven, no separate installation needed

Setup

STEP 1:

• Clone the Project:
• git clone
• cd selenium-mcp

Install Dependencies:Run the following command to download dependencies: mvn install

Configure the Server:Edit config.properties in the project root to set:

• server.port: HTTP server port (default: 8080)
• browser: Browser type (chrome or firefox, default: chrome)
• llm.endpoint: LLM API URL (default: http://localhost:11434 for Ollama)
• llm.model: LLM model name (default: llama3)
• llm.apiKey: API key for remote LLM (leave empty for Ollama)

Sample config.properties:

• server.port=8080
• browser=chrome
• llm.endpoint=http://localhost:11434
• llm.model=llama3
• llm.apiKey=

STEP 2:

• start server using McpServer.java
• Run selenium client file to see test execution SeleniumTestClient.java

Configure Selenium Operations:The selenium_operations.properties file defines supported Selenium commands and their parameters. It includes operations like start_browser, navigate, click_element, and more. Modify this file to add or change operations if needed. Example

• selenium_operations.properties:
• start_browser=browser,headless
• navigate=url
• click_element=by,value,timeout

... (other operations)

Set Up Ollama (if using local LLM):

Install Ollama (see Ollama documentation). Start Ollama server:

ollama serve

Ensure the specified model (e.g., llama3) is pulled:ollama pull llama3

Running the Server

Start the Server:Run the McpServer class:

mvn exec:java -Dexec.mainClass="McpServer"

Or, use your IDE to run McpServer.java. The server will start on the configured port (default: http://localhost:8080) and listen for POST requests to /execute.

Verify Server:Check the console for: MCP Server started on port 8080

Running the Client

Run Tests:Execute the TestNG client to send test steps and generate results: mvn test

Or, run SeleniumTestClient.java in your IDE with the TestNG plugin. The client sends a sample test prompt (e.g., navigating to https://example.com, clicking a link, and verifying text) and logs results using TestNG.

View Test Results:TestNG generates reports in the

... View full README on GitHub