The Agentic Shift: Building Zero-Code Flutter Integration Tests with Gherkin and Marionette

Traditional Flutter integration testing often involves writing and maintaining large amounts of test code that can become brittle as applications evolve. As AI agents become more capable, a new approach is emerging, one where developers describe expected behaviour in natural language and intelligent agents execute and validate those scenarios automatically. In this session, we'll explore how to build agentic integration testing workflows for Flutter applications by combining Gherkin(BDD) with the Marionette Model Context Protocol (MCP). You'll learn how to configure the Gherkin skill which utilises the Marionette MCP, enable AI agents to understand and execute Gherkin scenarios on real devices, and validate application state without writing conventional integration test code. Beyond demonstrating the workflow, we'll look at how teams can create reusable, codebase-specific skills that encapsulate common application behaviours, making AI-driven testing faster, more reliable, and easier to adopt across projects. Whether you're looking to reduce the maintenance burden of traditional integration tests or explore the next generation of AI-assisted developer workflows, this session provides practical techniques you can apply immediately. Learning Objectives - Learn how to install, configure, and use the Marionette MCP with AI agents for Flutter development. - Discover how Gherkin scenarios can drive fully agentic integration tests on real devices. - Build reusable, codebase-specific skills that streamline testing and automate common engineering workflows. Key Takeaways - Build and run a zero-code, AI-powered Flutter integration testing pipeline. - Design reusable agent skills that automate repetitive testing workflows. - Understand how MCP-enabled agents can improve the speed, reliability, and maintainability of Flutter integration testing.