Claude AI now learns your workflows by watching your screen, eliminating repetitive prompts

Claude AI Introduces Screen Recording to Teach Workflows

Anthropic has rolled out a significant update to Claude AI that fundamentally changes how users can automate tasks through its platform. The new feature, called "Teach Claude a skill" or "Record a skill," allows subscribers to demonstrate workflows through screen recordings rather than relying on written instructions1

1

2

2

. This screen-recorded teaching feature marks a shift toward more intuitive AI training methods, addressing a common frustration where AI chatbot systems repeatedly miss steps or execute tasks differently despite detailed prompts.

The functionality is integrated into Claude Cowork tools, available exclusively through the desktop app for paid subscribers3

3

. When users activate the record a skill feature, Claude AI observes everything: where you click, your keyboard inputs, mouse clicks, typing patterns, and any voiceover explanation you provide while performing the task1

1

4

4

. After processing the recording, the system converts your demonstration into what Anthropic calls "a repeatable skill" that can be executed on demand.

How the Feature Works for Repetitive Workflows

Accessing the capability requires users to open the Claude desktop app, navigate to Cowork, click the '+' button in the chat interface, and select "Record a Skill" from the dropdown menu under the Add files or photos section2

2

4

4

. Users then perform their task normally while Claude watches and learns. The feature targets repetitive workflows like organizing invoices in spreadsheets, renaming multiple image files, preparing weekly sales reports, or updating project trackers2

2

.

For instance, if you manually compile a regular report that pulls data from multiple sources, you can show Claude AI where to locate each piece of information and how you want it formatted. The system then creates a skill you can activate whenever you need that report generated1

1

. This approach eliminates the need for users to write lengthy prompts or create manual scripts, making automation accessible even to those without technical backgrounds4

4

.

Availability and Subscription Requirements

The feature began rolling out to users on Tuesday and is currently available exclusively to Claude Pro, Max subscribers, and Team subscribers1

1

2

2

3

3

. There's no indication that Anthropic plans to extend this capability to free Claude users. Multiple subscribers reported gaining access shortly after the announcement, with the feature appearing in their Cowork interface1

1

.

However, usage appears to have limitations. According to one Reddit user on the Pro plan, the option was available only once every five hours1

1

. Anthropic has not yet confirmed official limits on how many times users can record skills on each subscription tier, nor whether different recording lengths or task complexity affects the cooldown period. Users must also maintain a stable internet connection and use the Claude desktop app, as the feature is not currently available on mobile or web versions4

4

.

Positioning Claude as a Digital Teammate

Anthropic continues positioning Claude Cowork tools as more than just an AI assistant, aiming instead for the concept of a digital teammate that understands and adapts to individual work styles2

2

. This update builds on existing Cowork capabilities including file access, memory functions, connectors, and projects, creating a more comprehensive automation ecosystem4

4

. The ability to learn your workflows through observation rather than instruction represents a practical step toward AI systems that require less management overhead.

For professionals who perform identical tasks daily, this could prove transformative in reclaiming time previously spent on manual processes. The learned skills are saved in a user's skill library, allowing quick access whenever those tasks need repeating4

4

. As organizations increasingly seek ways to streamline operations without extensive technical implementation, features like this lower the barrier to meaningful automation. Watching your screen to understand context may signal where AI assistance is headed: less about following commands and more about observing and adapting to human work patterns.