Point at a button, say "fix this", Cursor's design mode finally speaks visual

Cursor has been quietly reshaping what an AI-powered IDE can do. The chat window is convenient, but design work is spatial. A developer sees a misaligned button, a designer spots a padding issue, a product manager notices a missing state, and the first instinct is to point, not type. The subtle trap waiting for AI agents in production

The company's latest update to Design Mode, announced today, turns that instinct into a workflow. Instead of translating a visual problem into a text prompt, users can click on any element, draw on the page, or describe the change by voice. The agent gets the context it needs to modify the code while the user moves on to the next fix. Cursor's team marketplaces get MCP servers and…

"The loop between what you spot and what you fix becomes shorter," the team wrote in a blog post. "You can designate precisely the part of the interface you're talking about without leaving the running product."

More than a prompt

The core idea is that a text instruction strips away too much information. A designer who says "move that card up" leaves the agent to guess which card, by how much, and relative to what. Design Mode avoids that ambiguity by giving the agent two complementary signals when an element is selected: the element's identity (xpath, component, attributes, computed styles, fiber tree props) and a screenshot for spatial context (layout, surrounding elements, exact page state).

Users can select multiple elements at once, useful when the change depends on the relationship between them. Drawing over the page is another option: circling a cluttered section, framing a zone, or marking a part of an animated page. The annotation overlays a frozen image of the viewport so the agent sees exactly what the user was reacting to.

Voice dictation is also new in this version. The combination means the user can point, draw, and speak to describe a modification, three modalities that together carry far more information than a text prompt alone. OpenAI's gpt-live-1 finally stops waiting its turn

Matching the model to the rhythm

Design Mode is built to handle the fast, iterative nature of frontend work. A developer adjusting one component will immediately notice the spacing around it, then see how another component should adapt. The new mode lets users queue up multiple changes: point at an element, describe the fix, move to another part of the page, and send another modification before the first one finishes. Multiple sub-agents run in parallel, each working on its assigned task.

This workflow works best with a model that can apply targeted changes quickly. Cursor recommends its Composer 2.5 model for the task, describing it as "both fast and highly effective for UI work." As agents finish, the application hot-reloads, changes appear in the running product instantly, and the user keeps iterating until the interface looks right. Anthropic Launches Claude Sonnet 5: A New Frontier in…

Closing the visual gap

Design Mode is part of a broader push inside Cursor to reduce the distance between what a user sees and what the agent understands. The company argues that the future of software development involves fluidly moving between higher levels of abstraction and lower-level details, all while staying in a state of flow.

"Design Mode gives users the control, autonomy, and precision editing tools that make this possible," the team wrote.

The update is available now in the agents window within Cursor's Browser. For developers who spend their days tweaking UIs, it is a small change with an outsize effect: the agent finally sees what they are pointing at. Meta's bet on Muse Spark is a bet that control, not…