Anthropic Releases Claude Update That Can Control Your Computer
Anthropic launches a new "Computer Use" API for its Claude 3.5 Sonnet model, allowing the AI to navigate desktop applications by mimicking mouse movements and keystrokes. This development brings the company closer to its goal of creating virtual assistants that automate everyday back-office tasks.
Anthropic releases an upgraded version of its Claude 3.5 Sonnet model that can understand and interact with desktop applications. Through a new "Computer Use" API available in open beta, the AI mimics human actions like keystrokes, button clicks, and mouse gestures to essentially act as a person sitting at a PC. The model analyzes screenshots to determine exactly where to move the cursor by counting pixels to execute tasks accurately.
This release fulfills a promise Anthropic makes to investors about building virtual assistants that handle research, emails, and other back-office jobs independently. Developers can access this computer-controlling capability through Anthropic’s API, Amazon Bedrock, and Google Cloud’s Vertex AI platform. The company also rolls out a standard version of the upgraded 3.5 Sonnet model to its Claude applications without the computer control feature.
The move places Anthropic in a crowded race to develop AI agents that automate software workflows. While tools for PC automation already exist from older RPA vendors and newer startups, Anthropic calls its approach an "action-execution layer" that provides a path toward monetizing massive AI investments. Major tech companies like Microsoft, Salesforce, and OpenAI are all pursuing similar agent technologies as organizational adoption continues to accelerate rapidly.