Designing VS Code UX for the Agentic Era
Visual Studio Codego watch the original →
the gist
The VS Code design team is evolving the editor for agentic workflows by prioritizing focused, minimal interfaces that help users manage multiple concurrent agents without becoming overwhelmed by feature bloat.
Balancing Agentic Complexity with Focus
The VS Code design team is navigating the transition to agentic workflows by shifting from feature-heavy interfaces to focused, purpose-built views. As users begin running multiple agents in parallel, the primary design challenge is preventing cognitive overload. The team actively prunes unnecessary UI elements, questioning every element's necessity to ensure it assists in decision-making rather than adding distraction. This approach is exemplified by the evolution of the Agents window, which was redesigned to provide a centralized space for monitoring sessions across different projects without cluttering the main editor window.
The Shift Toward Cross-Disciplinary Development
The boundary between design and engineering is blurring as the team adopts a model where everyone contributes to both code and design. Designers now frequently submit pull requests, while engineers leverage design guidelines and prototypes to iterate on features independently. This shift is supported by tools like Figma, which allows for bidirectional synchronization between design components and the codebase. While this increased breadth allows the team to move faster, the team maintains quality by relying on deep domain expertise in UX and visual design to provide the necessary "taste" that prevents AI-generated interfaces from becoming cluttered or unusable.
Validation and Accountability
To manage the velocity of AI-driven development, the team emphasizes rigorous validation loops over finality. Features are deployed to the Insiders build to gather telemetry and user feedback, allowing the team to iterate continuously. Data-driven decision-making is now more accessible to designers, who use Kusto queries to analyze telemetry without needing to rely on dedicated data analysts. The team maintains a culture of accountability by being willing to remove features that do not solve the intended problems, resisting the urge to simply add more functionality as AI capabilities expand.