Cursor 3.1 Launches Canvas Feature for Interactive AI Output Interfaces

ME News reports that on April 16 (UTC+8), according to monitoring by Beating, Cursor 3.1 has introduced the Canvas feature. After completing a task, the Agent no longer outputs only code or markdown text—it can now directly generate an interactive React interface containing charts, tables, diff views, or custom logic. Canvas exists as a persistent artifact within the Agents Window, displayed side by side with the terminal, browser, and version control. The core improvement lies in information density. For example, in incident response: previously, the Agent would compile time-series data from Datadog, Databricks, and Sentry into a markdown table, which was difficult to interpret and required manual visualization. Now, the Agent consolidates multi-source data directly into a single Canvas chart. The same applies to PR reviews: the Agent can group changes by importance and generate pseudocode for complex algorithms, rather than laying out all diffs flatly. The Cursor team themselves used Canvas in their Eval analysis. Previously, engineers had to manually review request IDs to identify failure patterns and even considered building a separate web app—ultimately, they created a Skill within Cursor that enables the Agent to ingest all rollout data, categorize failure causes, and generate an interactive analytical interface. This workflow significantly reduced troubleshooting time during the team’s last two model deployments. Canvas supports custom extensions: the Docs Canvas Skill available on the Cursor Marketplace allows the Agent to generate interactive architecture diagrams for any code repository. Users can also write their own Skills to define the type of Canvas the Agent generates. (Source: BlockBeats)