Oracolo del Test - Beyond Autocomplete: Architecting Black-Box Agentic Testing

Ciao Platipi,

iniziamo il 2026 con una serata a tema Agenti AI.
Mercoledi 30 Settembre presso la Sala vetrata Bassini, al piano terra di via Bassini 6 del Politecnico di Milano

Ludovico Besana ci parlerà di :

Beyond Autocomplete: Architecting Black-Box Agentic Testing

AI-assisted testing usually stops at generating test code. But what happens when AI agents can execute tests themselves?

This session explores Black-Box Agentic Testing, where AI agents interact directly with a running application through the Model Context Protocol (MCP). Instead of writing Playwright scripts, agents receive browser and testing tools and autonomously navigate interfaces, manipulate inputs, discover failures, and report findings using a Think → Act → Observe loop.

Through a live demo, we'll connect Playwright MCP and Chrome DevTools MCP to an LLM and watch it test a real application without traditional automation scripts.

We'll also examine the limitations of agentic testing: token costs, latency, non-deterministic execution, flaky outcomes, and the guardrails required to keep agents safe.

Attendees will learn where agentic testing fits in a modern quality strategy and how to start experimenting with MCP-powered exploratory testing.