Interact with agent-created visualizations in canvases

2026-04-15 · Source: Cursor Blog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Data Science & Analytics · Depth: Intermediate, short

Summary

Cursor, an AI code editor, now features agent-created canvases, enabling visual and interactive representation of information instead of text-heavy chat or markdown. Released on April 15, 2026, this functionality allows agents to generate custom dashboards and interfaces with tailored logic and interactivity for tasks like PR reviews, library learning, or agent management. Canvases are durable artifacts within the Agents Window, rendered using a React-based UI library with first-party components such as tables, diagrams, and charts, and can integrate existing Cursor components like diffs. Users can create skills, like the Docs Canvas skill, to teach agents how to generate specific canvas types, such as interactive architecture diagrams. Cursor uses canvases for data-intensive tasks like incident response dashboards, PR review interfaces, and eval analysis, significantly improving data interpretation and workflow efficiency.

Key takeaway

For AI Engineers and Machine Learning Engineers managing complex data or code reviews, Cursor's new canvas feature in version 3.1 offers a significant upgrade. You should explore using agent-created canvases to visualize eval results, consolidate observability data, or streamline PR reviews, as this can reveal hidden patterns and reduce manual effort compared to text-based analysis. Consider creating custom skills to tailor canvas generation to your specific workflows.

Key insights

Agent-created canvases transform text-based AI interactions into interactive, visual data explorations.

Principles

• Visualizations enhance data-intensive task comprehension.
• Interactive interfaces improve human-agent collaboration.

Method

Agents use a React-based UI library with components (tables, charts, diffs) to construct interactive canvases, following data visualization best practices, often integrating multiple data sources.

In practice

• Generate interactive architecture diagrams from codebases.
• Visualize time-series observability data from multiple sources.
• Group and prioritize changes in large code reviews.

Topics

• Agent-created Visualizations
• Cursor Canvases
• Incident Response Dashboards
• PR Review Interface
• Eval Analysis

Code references

• karpathy/autoresearch

Best for: AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• Cursor 3 Introduces Agent-First Interface, Moving beyond the IDE Model — Cursor 3 shifts AI coding from IDE-centric to an agent-first interface, reflecting evolving developer workflows.
• Agents on the Canvas in tldraw — Steve Ruiz, tldraw — AI agents can evolve from single-shot generation to collaborative, orchestrated entities directly interacting with canvas environments.
• Agentic Authoring of Interactive Multiview Visualizations in Genomics — Agentic LLM iteration significantly enhances genomics visualization quality, but complex agent architectures offer no further benefit.
• How we compare model quality in Cursor — Effective coding agent evaluation requires a hybrid online-offline approach aligned with real developer workflows.
• Build Interactive Agents with Generative UI — AI agents are evolving from text-only outputs to generating interactive, custom user interfaces.
• The man behind Cursor's "memory" feature — Cursor's Memory feature enhances agent context by distinguishing between task-specific and generalizable information for improved coding assistance.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Cursor Blog.