BrowserStack MCP Server: testing on real devices
BrowserStack MCP Server
The official BrowserStack MCP server: real devices, automated tests and test cases driven by an agent from your editor
- 154 ★ on the repository
- Last commit
Broad access
Why this level
- The agent starts paid sessions and tests on cloud devices and changes test case data under your account
- App builds and files go to the BrowserStack cloud when uploaded
Install
- Works in
Manual install
claude mcp add --scope user --transport http browserstack https://mcp.browserstack.com/mcpThen run /mcp in a session and sign in to BrowserStack.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent starts paid sessions and tests on cloud devices and changes test case data under your account
- App builds and files go to the BrowserStack cloud when uploaded
This is third-party code. Review the repository files before installing.
What it does
The server connects the BrowserStack testing platform to Cursor, Claude, VS Code and other MCP clients. From ordinary requests an agent opens mobile apps on real devices with the latest OS versions, reproduces crashes and investigates performance issues, opens sites in various browsers (including locally hosted ones) and runs Playwright and Selenium automated tests. For broken scripts the agent analyses logs and suggests fixes. Other scenarios are accessibility checks, test case management and load testing, plus access to BrowserStack's AI agents. To avoid filling the context with a tool per workflow, the server exposes five core ones: `listProducts`, `describeEntity`, `searchCapability`, `describeCapability` and `invokeCapability`. The agent finds the needed capability by description and runs it, and changes to data go ahead only with your approval. The hosted remote server `https://mcp.browserstack.com/mcp` is the most convenient: it needs no install and signs in through OAuth. A local run needs Node.js 22 or newer and a `username` and `access_key` pair.
Who it is for. QA engineers and developers who test sites and mobile apps on BrowserStack and want to drive tests from the editor in plain language.
Good fit when
- An app crashes on Android 14 and you need to open it on a real device and collect logs
- A site must be checked in Edge and several Chrome versions without installing them on your machine
- A Playwright test failed in CI and the agent must read the log and propose a fix
- You want to check page accessibility and manage test cases in one conversation
Not a fit when
- You must test apps behind a VPN, firewall or on localhost through the remote server: that needs the separate BrowserStack Local MCP
- You have no BrowserStack account: the server is unavailable without one, though open-source projects get a free plan
- Tests run on your own device fleet: the server works with the BrowserStack cloud
Example request
Open my site on the latest Edge and on an iPhone 15 Pro Max, take screenshots and show the console errors.MCP
- Transport
- http
- Authentication
- OAuth
Environment variables
MCP_UPLOAD_BASE_DIR- Folder from which uploading local files and apps is allowed.
Limitations
You need a BrowserStack account (open-source projects get a free plan). The remote server does not support Local Testing, and its latency is slightly higher. A local run needs Node.js 22 or newer. Uploading local files and apps is restricted to the folder in `MCP_UPLOAD_BASE_DIR`.
The author warns that the server is under active development and supports only a subset of the MCP spec. Calls go through the client and the model, so behaviour is sometimes unpredictable. The repository is published under AGPL-3.0, which matters if you plan to embed its code in a closed product.
How to disable. Remove the server with `claude mcp remove browserstack` or from the client settings. For the remote server, revoke access in your BrowserStack account settings.
FAQ
How do I connect BrowserStack MCP to Claude Code, Codex or Cursor?
Claude Code: `claude mcp add --scope user --transport http browserstack https://mcp.browserstack.com/mcp`, then `/mcp` and sign in. Codex: `codex mcp add browserstack --url https://mcp.browserstack.com/mcp` and `codex mcp login browserstack`. Cursor: add `"browserstack": { "url": "https://mcp.browserstack.com/mcp" }` to `~/.cursor/mcp.json`.
What can I test through this server?
Mobile apps on real devices, sites in various browsers, Playwright and Selenium automated tests, accessibility, test cases and load. The full list of tools by product is in the repository's docs/TOOLS.md.
Do I need my own Node install for the remote server?
No, the hosted server works with no local install or binaries and suits networks where local installation is not allowed. Node 22 or newer is needed only for a local run.
Do I need a local binary or is the remote server enough?
The remote one is enough: `https://mcp.browserstack.com/mcp` works with no install and signs in through OAuth. The limit is that it cannot do Local Testing (behind a VPN, firewall or on localhost), which needs the separate BrowserStack Local MCP.
What do the five core tools do?
`listProducts` lists the catalogue products, `describeEntity` explains an entity, `searchCapability` finds capabilities from a plain-language request, `describeCapability` gives parameters and `invokeCapability` runs a capability. Changes to data need your approval.
Discussion
No questions or reviews yet.
Related
MCP serversWebdriverIO MCP: browser and mobile app automation
webdriverio-mcp
The official MCP from the WebdriverIO team drives Chrome, Firefox, Edge, Safari, Electron, iOS and Android through one tool set
A community Playwright MCP server: browser automation, screenshots, device emulation and test code generation
An MCP server that spins up several browser agents to test a site for broken links, UI bugs and accessibility issues
kogiQA MCP gives an agent a browser that finds page elements without selectors, for debugging web apps and writing e2e tests
The SmartBear server gives the agent BugSnag, Swagger, Zephyr, QMetry and PactFlow: errors, API specs, test cases and contracts
CLI, MCP server and typed Node.js API so an agent can verify its changes in a live mobile app on iOS, Android and more
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| BrowserStack MCP Server: testing on real devices (this card) | yes | OAuth | 154 | |
| Appium MCP | yes | not required | 517 | drives a mobile app through Appium on your machine, with no real device cloud |
| Playwright CLI | yes | 15,133 | runs a browser locally without a cloud device fleet |
README in short
The README describes the server's purpose, quick setup in IDEs and assistants, a prompt example for an agent, a table of commands per client, examples for manual and automated testing, accessibility, test management and load, the five core tools and the limits of the remote server.