Skip to main content

WebdriverIO v10: Verification Loops for Coding Agents, on Every Platform

· 30 min read

WebdriverIO v10 is out today, and it is built for agent verification loops. Your coding agent can open your app in a real browser, on an Android or iOS device, or as a desktop app, check that the feature it just wrote actually works, and leave behind a regular test that keeps proving it on every change.

The highlights:

  • wdio session: a CLI that lets an agent drive browsers, phones and desktop apps step by step, then export what worked as a test.
  • browser.act(): write a step as intent, let a model work it out once, then replay it as plain code without spending a token.
  • WebdriverIO DevTools: a live debugger and portable traces, for you and for your agent.
  • Tabs, windows and frames as values you hold, instead of a hidden pointer you switch.
  • Breaking changes worth knowing: W3C-only sessions, Appium 3, Node.js 22.19+ and a strict $.

Upgrading an existing suite? Let your agent do it:

npx skills add webdriverio/webdriverio --skill wdio-v10-migration

Starting fresh? Run npm init wdio and say yes to coding agent support.

This WebDriver Went Tracing

· 5 min read

We've all been there. A test fails in CI. You open the logs, and all you get is:

Error: expected 'tomsmith' to be 'tomsmih'

No screenshot (maybe one at the end). No timeline. No idea what the browser was doing when it happened. Just a lonely assertion error and the slow realization that you'll be running this locally six times to reproduce it.

That's why we added trace mode to @wdio/devtools-service. It's like a black box recorder for your tests, with every navigation, click, keystroke, network request, console log, and DOM snapshot, written to disk so you can replay the whole run after the fact.

Introducing WebdriverIO MCP: We're Late, But We Brought Snacks

· 2 min read

Yes, we know. Playwright has an MCP server. Appium has one too. We're not first to the party.

But if you're already using WebdriverIO for cross-platform testing, you know it handles both web and mobile in one framework. Why should your AI assistant need two separate MCP servers to do the same thing?

That's why we built @wdio/mcp. One server for browsers and mobile apps, built on the WebdriverIO you already know.

Welcome! How can I help?

WebdriverIO AI Copilot