What's Coming Next
A look at where Certyn is heading — native app testing, device clouds, Chrome extensions, and the bring-your-own-environment model.
Certyn started as a way to test web apps with AI agents. Write a scenario in plain language, and an agent runs it in a real browser. That core loop isn't changing — but what the agent can test is about to get much wider.
Here's what we're working toward.
Stateful Environments
Today, every agent session starts from a blank browser. That's fine for testing public pages, but most real testing needs context — logged-in sessions, pre-configured tools, connected services.
We're building stateful environments. You set up the environment once — log in to Appetize.io, connect your BrowserStack account, authenticate with your staging server — and Certyn remembers it. Every future test session opens with that environment already prepared. No repeated logins, no wasted setup time, no credentials passed around in test instructions.
This is the foundation for everything else on this list.
Native Mobile App Testing
Our recent experiment with Appetize.io proved that the agent can interact with native iOS and Android apps streamed through the browser. The agent doesn't care whether it's looking at a webpage or a phone emulator — it sees a screen and interacts with it.
With stateful environments, you log in to your device cloud provider once, set up your app, and pick your target device. After that, every test session opens the provider directly — the device is ready, the app is loaded, and the agent starts testing immediately. No embed URLs, no setup overhead.
We're evaluating providers like Appetize.io, LambdaTest, and BrowserStack for this. The goal is zero setup per test run.
Chrome Extension Testing
Chrome extensions are notoriously undertested. Most QA tools ignore them entirely.
Certyn can already load and test published extensions from the Chrome Web Store. You provide the extension URL, the agent installs it, and tests the popup, content scripts, options page — whatever you describe in your test case.
We're working on making this a first-class workflow with dedicated support for extension-specific scenarios like toolbar interactions, permission prompts, and cross-tab behavior.
Bring Your Own Environment
Right now, Certyn runs agents in our managed containers. But the underlying architecture doesn't require it. The agent just needs a screen to look at and tools to interact with it.
We already support custom agent images — you can specify your own Docker image with any environment you need. This opens the door to testing things like:
- Electron apps — they're Chromium-based, so the agent connects via remote debugging
- Kiosk interfaces — embed a VNC viewer and point the agent at it
- Any device cloud — connect to BrowserStack, Sauce Labs, Samsung Remote Test Lab, or AWS Device Farm through their web interfaces
We're not building integrations for every platform. We're building a flexible agent that works with whatever you put in front of it.
Better Environment Simulation
We already support device emulation, network throttling, timezone, geolocation, and dark mode. We're expanding this with:
- More device presets — covering newer phones, foldables, and common tablet sizes
- Custom network profiles — define your own latency and throughput numbers
- Combined scenarios — test on iPhone 14 with slow 3G in the Tokyo timezone, all in one instruction
Performance and Accessibility as First-Class Citizens
The agent already collects performance metrics and can inspect the accessibility tree. We're turning these from tools you can request into automatic checks that run alongside every test — Core Web Vitals captured by default, accessibility violations flagged without being asked.
What We're Not Building
- Cross-browser testing — Chromium covers the vast majority of real-world usage. Firefox and Safari support isn't on the near-term roadmap.
- Visual regression diffing — the agent takes screenshots and can compare them, but pixel-level diffing tools already exist. We'd rather integrate than rebuild.
- Load testing — Certyn tests behavior, not throughput. We're not competing with k6 or Locust.
The Direction
The pattern is consistent: instead of building specialized infrastructure for every testing scenario, we're making the agent flexible enough to work with existing tools and services. Device clouds, extension stores, custom environments — the agent adapts to what's already there.
If it has a screen, Certyn can test it.