BrowserStack MCP Server – Cloud Testing Tools
BrowserStack's official MCP Server connects AI coding assistants to BrowserStack's testing infrastructure and products. Use it to launch manual sessions, run automated web and mobile tests, create and manage test cases, retrieve failure logs, run accessibility scans, and perform visual testing without leaving the IDE or chat client.
Overview
BrowserStack MCP Server connects AI agents and LLM-powered development tools to
BrowserStack's real device cloud and testing products. It lets developers and
QA teams use natural-language prompts to run, debug, and manage testing
workflows from tools such as VS Code, Cursor, Claude, Cline, and other
MCP-compatible clients.
What the MCP server enables
BrowserStack MCP exposes tools and workflows across BrowserStack products. An AI
agent can:
- Launch BrowserStack Live sessions for interactive cross-browser testing.
- Launch App Live sessions for manual mobile app testing on real devices.
- Set up and run BrowserStack Automate tests for web applications.
- Run App Automate workflows for mobile app automation and screenshots.
- Create, manage, and retrieve test cases in Test Management.
- Retrieve failure logs and analytics from Test Reporting & Analytics.
- Run accessibility scans and get contextual accessibility guidance.
- Perform Percy visual testing and UI review workflows.
- Use AI-agent workflows to generate test cases, debug failures, and suggest
code fixes from testing context.
When to use it
Use BrowserStack MCP when an AI coding assistant needs to validate an app across
real browsers, devices, and platforms. Practical examples include launching a
Live session to reproduce a Safari issue, setting up BrowserStack SDK support
for a Selenium suite, running an App Automate test on a real Android device,
retrieving screenshots from a failed session, scanning a page for accessibility
issues, or creating test cases from a feature specification.
Connection and authentication
BrowserStack supports two setup paths. The remote MCP server is available at
https://mcp.browserstack.com/mcp, requires no local installation, and uses
OAuth to authenticate securely with BrowserStack. BrowserStack recommends this
approach when a client supports HTTP-based remote MCP.
The local MCP server runs on the user's machine with
npx -y @browserstack/mcp-server@latest. Local setup requires
BROWSERSTACK_USERNAME and BROWSERSTACK_ACCESS_KEY, and BrowserStack
documents Node.js v22+ as a prerequisite. Local mode is useful when testing
local development or staging environments, including sites or mobile apps behind
a corporate firewall.
Key considerations
Remote MCP reduces manual credential handling because it uses OAuth. Local MCP
keeps BrowserStack credentials and sensitive project details on the user's
machine, but credentials must still be stored securely as environment variables
rather than plain text where possible. BrowserStack testing actions can start
sessions, run automated tests, access session artifacts, and create or modify
testing data, so review prompts before allowing costly or state-changing
workflows. Product availability depends on the user's BrowserStack plan,
permissions, enabled products, and available real devices, browsers, or
automation minutes.
Supported Transports
streamable_http
URL: https://mcp.browserstack.com/mcp
stdio
Command: npx
Args:
-y@browserstack/mcp-server@latest
Frequently Asked Questions
- When should an AI agent use the BrowserStack MCP Server?
- Use it when an agent needs to validate, debug, or manage testing workflows on BrowserStack, such as launching Live sessions, running web or mobile automation, creating test cases, retrieving failure logs, scanning for accessibility issues, or running Percy visual testing.
- What does the BrowserStack MCP Server add to an AI agent's capabilities?
- It gives the agent access to BrowserStack's real-device cloud and testing products through MCP tools, enabling natural-language workflows for manual testing, automated testing, test management, accessibility, visual testing, and failure analysis directly from an AI-enabled client.
- What can an AI agent access or manage through BrowserStack MCP?
- Depending on the user's BrowserStack plan and permissions, the agent can use tools for BrowserStack Live, App Live, Automate, App Automate, Test Management, Accessibility, Test Reporting & Analytics, Percy visual testing, and AI-agent-assisted test generation or debugging workflows.
- How is authentication configured for BrowserStack MCP?
- Remote MCP authenticates through OAuth at https://mcp.browserstack.com/mcp. Local MCP uses BrowserStack API-key authentication with BROWSERSTACK_USERNAME and BROWSERSTACK_ACCESS_KEY. Store local credentials as environment variables and avoid committing access keys to configuration files.
- Which transport should be used for BrowserStack MCP?
- Use the remote Streamable HTTP endpoint when the client supports it and you want OAuth with no local installation. Use local stdio with `npx -y @browserstack/mcp-server@latest` when testing local, staging, or firewall-protected environments or when the client requires a local server.