Agentic Workflows and Context Awareness

Chrome DevTools has evolved from simple turn-based chat into an agentic debugging environment. Instead of requiring manual selection of DOM elements or network requests, the AI assistant now uses a context-selection agent to automatically identify relevant data. For instance, if you ask to improve page performance, the agent autonomously reloads the page, records a performance trace, and analyzes Core Web Vitals (LCP, INP, CLS) without further user intervention. This shift has reduced the word count of AI responses by 68% on average, prioritizing concise, actionable insights over verbose explanations.

Transparent Reasoning with Structured Walkthroughs

To ensure developers remain in control, the updated AI assistant includes "Agent Walkthroughs." When Gemini performs an investigation, it generates a step-by-step record displayed in a split-view panel. This interface uses specialized widgets to visualize findings—such as computed styles, DOM trees, or Lighthouse audit scores. Each widget includes a "reveal" button that links directly to the corresponding data source within DevTools, allowing developers to verify the AI's logic and replicate the investigation manually.

Integration with External Coding Agents

Recognizing that debugging often happens in the browser but implementation happens in the IDE, Chrome introduced a "Copy for your coding agent" feature. This button condenses the entire investigation, including the context and potential fixes, into a structured prompt. This allows developers to seamlessly hand off findings to agentic IDEs like Cursor or Google Antigravity. Additionally, the new WebMCP (Model Context Protocol) support in the Application panel allows developers to inspect and debug how websites expose specialized tools to AI agents, ensuring the browser remains compatible with the evolving agentic web ecosystem.