What is TestingBot MCP?

What you can do

The TestingBot MCP server ships as two bundled packages that together expose a broad set of tools. They fall into two broad areas.

Account & test management

Query and manage everything in your TestingBot account through conversation, without opening the dashboard:

• Tests: list recent tests, read full session details, update metadata, mark passed/failed, stop or delete tests.
• Builds: list builds, see the tests inside a build, and delete builds.
• Browsers & devices: list available desktop browsers, platforms and mobile devices (real and emulators/simulators).
• Screenshots: capture a URL across multiple browsers and retrieve the resulting images.
• Storage: upload native mobile apps (.apk/.ipa/.zip) for testing.
• Team & user: view account info, team settings and usage, and update your profile.
• Tunnels & logs: list/stop tunnels and fetch session failure logs for debugging.
• Project scaffolding: scan a project for test files and generate a ready-to-paste TestingBot config for your framework.

Live browser & mobile automation

Beyond account management, the bundled automation package lets the agent drive real sessions step by step:

• Desktop browsers: open real Chrome, Firefox, Edge, Safari or IE on Windows, macOS and Linux, then navigate, click, type, read text, run JavaScript and take screenshots.
• Mobile browsers: drive Chrome/Safari on Android and iOS, on emulators/simulators (the cheaper default) or on physical real devices.
• Native mobile apps: launch and control uploaded iOS/Android apps with Appium.
• Interactive live sessions: start a manual live testing session, or watch and take over any automated session via its live-view URL.
• CDP sessions: get a remote browser with a Chrome DevTools Protocol endpoint to drive with Playwright or Puppeteer.

For the complete list of tools and their parameters, see the Tool reference. For end-to-end examples, see Use cases & workflows.

How it works

The MCP server runs as a local process, started by your AI client over standard input/output. When you type a request, your AI client decides which tools to call and sends those calls to the server. The server then does one of two things:

1. For account and test operations, it makes authenticated calls to the TestingBot API and returns the results to your client in a readable format.
2. For automation, it spins up a real cloud session (desktop or mobile) and issues WebDriver/Appium commands against it, returning a sessionId and a live-view URL so you can follow along.

The MCP server is distributed as @testingbot/mcp-server, which handles account and orchestration tools and bundles @testingbot/automation-mcp for live browser and mobile automation. You install just the one package, both sets of tools become available automatically.

Every automation session returns a live-view URL in the form https://testingbot.com/tests/<sessionId>/live?auth=<hash> so you can watch the session run, or step in manually.

Supported clients

The TestingBot MCP server works with any MCP-compatible client through a standard stdio configuration. Tested clients include:

• Claude Desktop and Claude Code
• Cursor
• VS Code (Copilot / MCP)
• Windsurf
• Any other MCP-compatible client (generic stdio config)

1-click installers are available for VS Code and Cursor, and Claude can install the packaged .mcpb bundle. See the cards in the Quick start below, or the full Setup & installation guide for per-client configuration.

Quick start

Getting up and running takes three steps.

1. Install

Use one of the 1-click installers below, or add the server to your client's MCP configuration manually. The server runs through npx, so there is nothing to install globally:

npx -y @testingbot/mcp-server@latest

In your MCP client, that maps to the standard configuration block:

{
  "mcpServers": {
    "testingbot": {
      "command": "npx",
      "args": ["-y", "@testingbot/mcp-server@latest"],
      "env": {
        "TESTINGBOT_KEY": "<key>",
        "TESTINGBOT_SECRET": "<secret>"
      }
    }
  }
}

For 1-click setup, pick your editor below:

For per-client configuration file locations and step-by-step instructions for every supported client, see the full Setup & installation guide.

2. Authenticate

Once the server is configured, connect it to your TestingBot account. The recommended way is the tb_login tool, just ask your agent to log in, and a browser window opens for you to authorize. Nothing is pasted, and your credentials are written to ~/.testingbot/credentials on your machine and used automatically afterwards.

Alternatively, set TESTINGBOT_KEY and TESTINGBOT_SECRET in the client's env block (shown in the config above), useful for CI and shared configs. Get your keys from your account settings. See Authentication for browser login, device-code login for remote/headless machines, and credential storage details.

3. Ask your agent

That is it. Describe what you want in plain language and the agent will pick the right tools. For example:

"Open Chrome on Windows 11, go to https://example.com, and tell me the page title."

"List my last 10 TestingBot tests and show me the details of any that failed."

Browse many more end-to-end examples in Use cases & workflows.

Documentation index

Dive deeper into any part of the TestingBot MCP server:

• Setup & installation: prerequisites, install methods and per-client configuration for Claude, Cursor, VS Code, Windsurf and more.
• Authentication: browser login, device-code flow, API keys, credential storage and security.
• Tool reference: every tool, grouped by category, with full parameter tables.
• Use cases & workflows: real scenarios with example prompts and the tool sequences behind them.
• Troubleshooting: fixes for common issues with login, devices, screenshots and sessions.
• FAQ: answers to the most common questions about the MCP server.