Cursor Introduces Shared Canvases and /loop Skill for Collaborative AI Agents

TL;DR

• Cursor now allows teams to share interactive AI-generated "canvases" via a browser link.
• These shared canvases provide read-only snapshots of agent-created artifacts like reports and dashboards.
• A new "/loop skill" enables automated, long-running agent tasks on a local schedule.
• Both features are aimed at enhancing team collaboration and agent utility.

Cursor has significantly enhanced its collaborative capabilities with the introduction of Shared Canvases. Previously, sharing complex AI agent outputs meant sending entire chat threads, which could be cumbersome. Now, users can generate interactive artifacts, referred to as "canvases"—which can range from detailed reports and dashboards to entirely custom interfaces—and share them directly with their team. This new functionality allows for much more efficient dissemination of AI-generated work products.

These shared canvases are not just static images; they are live snapshots that can be accessed through a simple link in the browser. Team members can view these artifacts without needing to be logged into Cursor or have the specific agent context. The canvases are viewable within the Cursor Dashboard, providing a centralized hub for team-shared AI creations. This feature is a substantial step towards making AI outputs more accessible and actionable for broader teams.

Currently, the Shared Canvases feature is available to users on the Pro, Teams, and Enterprise plans. This tiered approach suggests that the feature is designed to be a premium offering, catering to professional development teams and organizations that rely heavily on AI-assisted workflows. More details on how to utilize this feature can be found in the documentation.

Complementing the shared canvases, Cursor has also rolled out the /loop skill. This innovative addition empowers users to run a specific prompt repeatedly on a local schedule. This is particularly useful for agents designed to perform long-running tasks, such as monitoring a system, continuously updating a report, or working on a feature until a specific condition is met. The /loop skill allows for automated, sustained agent activity without constant manual intervention.

The flexibility of the /loop skill is a key highlight. Users can set it to run at fixed intervals, ensuring a task is performed regularly, for example, checking deployment status every five minutes. Alternatively, if no fixed interval is specified, the agent can intelligently decide when to activate based on certain outcomes or when the user manually stops the process. This adaptive capability makes it suitable for a wide array of automated workflows, moving beyond simple scheduled tasks.

This new skill opens up possibilities for more sophisticated local long-running agents. Imagine an agent that continuously refines code until all tests pass or a bot that monitors a live feed for specific events. The /loop skill provides the underlying mechanism to enable such persistent, goal-oriented AI operations directly within the user's development environment. To explore other recent enhancements, users can visit the general Cursor Changelog.

Summary

• Cursor now enables teams to share interactive agent-generated canvases via browser links, offering read-only snapshots.
• The new /loop skill allows for automated, long-running agent tasks on a local schedule, enhancing workflow efficiency.
• Shared Canvases are available on Pro, Teams, and Enterprise plans.
• These updates aim to improve collaboration and the utility of AI agents for development teams.

Source: Shared Canvases and /loop Skill · Cursor