Anthropic Gives Claude 3.5 Sonnet a Mouse and Keyboard
The Computer Use API lets the model move a cursor, click, type, and run desktop software through screenshots and coordinate outputs. Developers call the API to execute full workflows such as form filling and multi-app sequences without hand-coding each click. The feature is live for all Claude 3.5 Sonnet users via the Anthropic SDK.
You stop writing brittle scripts for every button and instead let the model observe the screen state and act. This shifts your workflow from brittle automation to goal-level instructions that survive UI changes. The result is faster prototyping and fewer lines of glue code.
Anthropic released the API publicly and reported that early partners completed end-to-end accounting tasks in under three minutes with zero custom UI selectors. The same pattern is now being tested by solo founders at Y Combinator’s current batch.
Step 1: Sign up at https://console.anthropic.com and generate an API key. Step 2: Install the Anthropic Python SDK and call client.beta.computer with a prompt that describes the task and screenshot input. Step 3: Receive coordinate and keypress actions; the model executes them on a controlled desktop and returns the final screen state.