Anthropic has introduced a major upgrade to its AI system Claude, adding agentic capabilities that allow it to independently use tools, complete tasks, and function more like a personal assistant with minimal human input.
In addition, the update allows Claude to operate design platforms such as Figma and Canva directly from mobile devices, making it possible to complete professional work without needing a laptop.
Claude’s New Agentic Upgrade Turns It Into a Mobile-First AI Assistant
Furthermore, users can now assign tasks to Claude and let it carry them out autonomously, even when working from their smartphones, significantly improving convenience and accessibility.
The company has also emphasized that this upgrade enables Claude to independently execute workflows, reducing the need for constant human supervision while still maintaining control when necessary.
Alongside this, Anthropic shared the update publicly through a post on X, highlighting the expanded capabilities and demonstrating how the system works in real scenarios.
For example, Claude can now analyze Amplitude dashboards and interpret designs in Figma using simple prompts, showing its ability to handle complex professional tasks with ease.
Moreover, it can generate presentations through Canva without requiring users to exit the conversation, effectively merging multiple tools and workflows into a single mobile interface.
As a result, this integration allows users to complete a wide range of tasks in one place, especially when they are away from their computers.
At the same time, the system includes safety mechanisms within its “auto mode,” where an AI classifier evaluates the level of risk associated with each task before execution.
Specifically, this safety layer prevents harmful actions such as executing malicious code, deleting large amounts of data, or compromising sensitive information.
If a task is identified as risky, Claude will pause the process and request user approval before continuing, ensuring controlled autonomy.
Building on this, the newly introduced “auto mode” enables Claude to perform tasks without repeatedly asking for permission, particularly useful in workflows like coding.
Consequently, this reduces interruptions and streamlines task execution while still maintaining safeguards.
Claude’s Safer Auto Mode Replaces Unchecked AI Actions
The company also positions this feature as a safer alternative to options that completely bypass user permission, where AI systems act without any checks.
In one of its tweets, Anthropic stated:
“Your work tools in Claude are now available on mobile.
Explore Figma designs, create Canva slides, check Amplitude dashboards, all from your phone.
Give it a try: http://claude.com/download
Beyond design and analytics tools, Claude Code is capable of handling a wide variety of tasks, supported by its built-in risk assessment system that filters actions before execution.
Additionally, only actions deemed safe are ultimately carried out, reinforcing the platform’s focus on secure automation.
If unsafe actions are detected during processing, the system intervenes by stopping the task and asking the user for confirmation before proceeding.
Moving further, Claude now functions as a personal assistant by integrating with apps like Slack and Calendar, allowing it to perform tasks on behalf of the user.
It can also access emails, draft replies, and even compile weekly reports, reducing manual workload in everyday professional tasks.
