LocatorLabs MCP Server
# LocatorLabs MCP Server
> šÆ Intelligent Playwright locator generation powered by AI
An MCP (Model Context Protocol) server that provides smart locator generation for Playwright test automation.
**Works directly in VS Code** (with GitHub Copilot), Cursor, Windsurf, Cline, and Claude Desktop. No extra apps needed - just your favorite IDE!
[](https://www.npmjs.com/package/locatorlabs-mcp)
[](https://www.npmjs.com/package/locatorlabs-mcp)
### ā” Quick Install


**VS Code (v1.99+)** - Copy & run in terminal:
```bash
code --add-mcp '{"name":"locatorlabs","command":"npx","args":["-y","locatorlabs-mcp"]}'
```
**Cursor:**
```bash
cursor --add-mcp '{"name":"locatorlabs","command":"npx","args":["-y","locatorlabs-mcp"]}'
```
**Windows PowerShell:**
```powershell
code --add-mcp '{\"name\":\"locatorlabs\",\"command\":\"npx\",\"args\":[\"-y\",\"locatorlabs-mcp\"]}'
```
## ⨠Features
| Feature | Description |
|---------|-------------|
| šÆ **Get Locators** | Get all possible locators for any element (Playwright + Selenium formats) |
| š **Analyze Page** | Scan pages and identify all interactive elements |
| šļø **Generate POM** | Auto-generate Page Object classes for **Playwright** & **Selenium** |
| ā
**Run Tests** | Execute Playwright tests and get pass/fail results |
| š **Generate Tests** | Create executable test scripts |
## š§ Supported Frameworks
| Framework | Locators | Page Object Model |
|-----------|----------|-------------------|
| **Playwright** | ā
`getByRole`, `getByTestId`, etc. | ā
TypeScript, JavaScript, Python |
| **Selenium** | ā
`By.id`, `By.xpath`, `By.cssSelector` | ā
Java, Python, C# |
## š Quick Start
### For VS Code (v1.99+ with GitHub Copilot)
**One-Line Install (Mac/Linux):**
```bash
code --add-mcp '{"name":"locatorlabs","command":"npx","args":["-y","locatorlabs-mcp"]}'
```
**One-Line Install (Windows PowerShell):**
```powershell
code --add-mcp '{\"name\":\"locatorlabs\",\"command\":\"npx\",\"args\":[\"-y\",\"locatorlabs-mcp\"]}'
```
**Or Manual Setup:**
1. Open Command Palette (`Cmd+Shift+P` / `Ctrl+Shift+P`)
2. Type: `MCP: Add Server`
3. Select: `Command (stdio)`
4. Command: `npx`
5. Args: `-y locatorlabs-mcp`
**Using with GitHub Copilot:**
1. Enable Agent Mode: Settings ā Search `chat.agent.enabled` ā Enable
2. Open Copilot Chat (`Ctrl+Cmd+I` / `Ctrl+Alt+I`)
3. Switch to **Agent** mode (dropdown at top)
4. Chat: "Get locators for login button on https://saucedemo.com"
### For Claude Desktop
**Step 1:** Open config file
- **Mac:** `~/Library/Application Support/Claude/claude_desktop_config.json`
- **Windows:** `%APPDATA%\Claude\claude_desktop_config.json`
**Step 2:** Add this configuration:
```json
{
"mcpServers": {
"locatorlabs": {
"command": "npx",
"args": ["-y", "locatorlabs-mcp"]
}
}
}
```
**Step 3:** Restart Claude Desktop
### For Cursor IDE
**One-Line Install:**
```bash
cursor --add-mcp '{"name":"locatorlabs","command":"npx","args":["-y","locatorlabs-mcp"]}'
```
**Or Manual:** Go to Cursor Settings ā MCP ā Add new MCP Server
- Name: `locatorlabs`
- Command: `npx -y locatorlabs-mcp`
### For VS Code + Cline Extension
Add to Cline MCP settings:
```json
{
"mcpServers": {
"locatorlabs": {
"command": "npx",
"args": ["-y", "locatorlabs-mcp"]
}
}
}
```
## š Usage Examples
Just chat naturally:
### Get Locators for an Element
> "Get me all locators for the login button on https://www.saucedemo.com"
**Response:**
| Type | Locator | Reliability |
|------|---------|-------------|
| testId | `getByTestId('login-button')` | 98% |
| role | `getByRole('button', { name: 'Login' })` | 95% |
| id | `locator('#login-button')` | 90% |
### Analyze Entire Page
> "Analyze all form elements on https://www.saucedemo.com"
### Generate Page Object Model
**Playwright:**
> "Generate a TypeScript Page Object for https://www.saucedemo.com and call it LoginPage"
**Selenium:**
> "Generate a Java Selenium Page Object for https://www.saucedemo.com and call it LoginPage"
> "Generate a C# Selenium Page Object for https://www.saucedemo.com called LoginPage"
> "Generate a Python Selenium POM for https://www.saucedemo.com named LoginPage"
### Run a Test
> "Run a test that logs into saucedemo.com with standard_user and secret_sauce, then verify Products page appears"
**Response:**
```
ā
Test PASSED (3.2s)
- Navigate to login page ā
- Enter username ā
- Enter password ā
- Click login ā
- Verify Products visible ā
```
### Generate Test Script
> "Generate a Python test script for the saucedemo login flow"
## š ļø Available Tools
| Tool | Description |
|------|-------------|
| `get_locators` | Get all possible locators (Playwright + Selenium formats) |
| `analyze_page` | List all interactive elements on a page |
| `generate_page_object` | Create POM class (Playwright: TS/JS/Python, Selenium: Java/Python/C#) |
| `run_test` | Execute tests in real browser, get pass/fail |
| `generate_test` | Generate executable test scripts |
## š¦ Supported Languages for Page Objects
| Language | Framework | Example Output |
|----------|-----------|----------------|
| `typescript` | Playwright | `page.getByRole('button')` |
| `javascript` | Playwright | `page.getByRole('button')` |
| `python` | Playwright | `page.get_by_role('button')` |
| `java-selenium` | Selenium | `@FindBy(id = "btn")` |
| `python-selenium` | Selenium | `By.ID, "btn"` |
| `csharp-selenium` | Selenium | `[FindsBy(How = How.Id)]` |
## šÆ Supported Test Actions
| Action | Description |
|--------|-------------|
| `navigate` | Go to URL |
| `click` | Click element |
| `fill` | Enter text |
| `clear` | Clear input field |
| `check` / `uncheck` | Toggle checkbox |
| `select` | Select dropdown option |
| `hover` | Mouse hover |
| `press` | Keyboard key press |
| `assert_visible` | Verify element visible |
| `assert_hidden` | Verify element hidden |
| `assert_text` | Verify text content |
| `assert_value` | Verify input value |
| `assert_url` | Verify page URL |
| `assert_title` | Verify page title |
| `wait` | Wait for time |
| `wait_for_element` | Wait for element |
| `screenshot` | Capture screenshot |
## š Locator Priority
LocatorLabs ranks locators by reliability:
1. **data-testid** (98%) - Best, explicitly for testing
2. **Role + Name** (95%) - Playwright recommended
3. **Label** (90%) - Accessible
4. **ID** (90%) - Stable if meaningful
5. **Placeholder** (85%) - Good for inputs
6. **Text** (75%) - May change
7. **CSS** (60%) - Can be brittle
8. **XPath** (40%) - Avoid unless necessary
## š LocatorLabs vs Playwright MCP
| Feature | Playwright MCP | LocatorLabs MCP |
|---------|----------------|-----------------|
| Get all locators for element | ā | ā
|
| Locator reliability ranking | ā | ā
|
| Selenium locator format | ā | ā
|
| Generate Playwright POM | ā | ā
|
| Generate Selenium POM (Java/C#/Python) | ā | ā
|
| Run tests with pass/fail | ā | ā
|
| Browser automation | ā
| ā
|
| Device emulation | ā
| ā |
**They work great together!** Use Playwright MCP for navigation/scraping, LocatorLabs MCP for locators/POMs/tests.
## š Requirements
- Node.js 18+
- Playwright browsers (auto-installed on first run)
## š§ Troubleshooting
### Playwright browsers not installed
```bash
npx playwright install chromium
```
### Permission issues on Mac
```bash
chmod +x ~/.npm/_npx/*/node_modules/.bin/locatorlabs-mcp
```
### View logs (Claude Desktop)
```bash
tail -f ~/Library/Logs/Claude/mcp*.log
```
### Auto-approve tools in VS Code
Add to `settings.json`:
```json
{
"chat.tools.autoApprove": true
}
```
## š¤ Contributing
Contributions welcome! Please open an issue or PR on GitHub.
## šŗ Author
**Naveen AutomationLabs**
- YouTube: [@naveenautomationlabs](https://www.youtube.com/@naveenautomationlabs) (500K+ subscribers)
- Website: [naveenautomationlabs.com](https://naveenautomationlabs.com)
- LinkedIn: [Naveen AutomationLabs](https://linkedin.com/in/naveenkhunteta)
## š License
MIT
---
TDQS
Scored across 5 tools
Each tool has a distinct, non-overlapping purpose: analyze_page for page-wide element analysis, generate_page_object for POM generation, generate_test for test script creation, get_locators for element-specific locator retrieval, and run_test for test execution. The descriptions clearly differentiate their scopes (e.g., get_locators is for specific elements vs. analyze_page for entire pages), eliminating ambiguity.
All tool names follow a consistent verb_noun pattern with snake_case (e.g., analyze_page, generate_page_object, run_test). The verbs are action-oriented and descriptive (analyze, generate, get, run), creating a predictable and readable naming convention throughout the set.
With 5 tools, the server is well-scoped for its purpose of web automation and testing support. Each tool serves a unique function in the workflow (analysis, POM generation, test creation, locator retrieval, test execution), and none feel redundant or missing, making the count appropriate for the domain.
The toolset covers core web automation workflows comprehensively: analysis, POM generation, test creation, locator retrieval, and test execution. A minor gap exists in lacking direct element interaction tools (e.g., click_element or fill_form), but agents can work around this by using generated tests or locators, and the overall surface supports end-to-end testing tasks effectively.