Best for
- Web developers debugging and testing complex web applications
- QA engineers automating end-to-end testing
- AI agents needing selector-free browser interaction
K
kogiQA MCP
Community
Selector-free browser automation MCP server for AI agents, focused on debugging and testing.
GitHub source repository ↗
38FMRS · D
The kogiQA MCP Web Browser is a practical browser automation server that enables AI agents to interact with web pages efficiently through natural language and without needing selectors. It is particularly suited for development and testing scenarios, but requires attention to runtime environment and security risks.
The kogiQA MCP Web Browser is a Model Context Protocol (MCP) server that provides browser automation capabilities using the kogiQA browser control algorithm. It allows LLMs to interact with web pages through natural language, bypassing the need for screenshots or visually-tuned models, and without requiring selectors, saving time and tokens. It is designed for debugging style issues, automatically fixing console errors, mapping and documenting web app functionality, automatically writing end-to-end tests, and automating exploratory testing.
Tools
browser_navigate
Navigate to a specified URL.
browser_act
Perform actions on the page (e.g., click, type, hover).
browser_extract
Extract structured data or text from the page.
browser_evaluate
Execute JavaScript code in the page context.
browser_console
Retrieve or clear browser console logs.
browser_screenshot
Take a screenshot of the current page (for visual debugging).
Setup
- Ensure Node.js 20 or later is installed.
- In an MCP-capable client (such as Claude Desktop, VS Code, Cursor), add the server using npx:
npx kogiqa-mcp@latest. - For VS Code, use the provided buttons or CLI command; for Cursor, add the MCP server in settings with the above command.
- For other clients, add the provided JSON configuration to your MCP settings.
claude_desktop_config.json
{
"mcpServers": {
"kogiqa-browser": {
"command": "npx",
"args": [
"kogiqa-mcp@latest"
]
}
}
}
Fit and risk
Not for
- Tasks requiring high-precision visual verification (e.g., image recognition)
- Bypassing browser security restrictions
- Deployment in environments without Node.js 20+
Required permissions
- Requires Node.js runtime
- Requires network access to navigate web pages
- May require access to local development servers
- Requires ability to execute JavaScript code
Risks and side effects
- May execute unverified navigation or actions, leading to unexpected behavior
- May expose sensitive information (e.g., console logs)
- Depends on a third-party service (kogiQA)
- Potential security risks if browser automation is misused
Troubleshooting
- Check that Node.js version is 20 or higher
- Verify MCP client configuration with command 'npx kogiqa-mcp@latest'
- Check network connectivity to reach target web pages
- Inspect client logs for error messages
- Ensure target application is running at an accessible address
Use cases
Smoke-test a user flow: open a local app, log in, add a product, and verify cart badge.
Debug a style issue: check element overflow on a specific viewport and suggest CSS fixes.
Fix console errors: navigate through all pages, collect console errors, and fix them in source.
Map and document your app: explore internal links and produce a Markdown document describing each page.
Generate end-to-end tests: write Cypress tests for sign-up flow covering happy path and validation errors.
Exploratory testing: perform 10 minutes of exploratory testing, trying edge-case inputs and reporting issues.
Supported clients
Claude DesktopFull support
VS CodeFull support
CursorFull support