Claude 3.5 Sonnet Learns to Drive Your Desktop
Anthropic added a computer use API to Claude 3.5 Sonnet. The model receives screenshots, outputs mouse coordinates and keystrokes, and can open, edit, and save files across standard desktop applications. Early tests show the agent completing multi step workflows such as spreadsheet analysis in under two minutes.
Plain language instructions now replace brittle scripts for routine UI tasks. Users can delegate repetitive desktop operations to the model and focus attention on higher level decisions rather than writing or debugging automation code.
The Anthropic team published benchmark results showing a 20 percent success rate on OSWorld tasks without task specific fine tuning. Several freelance analysts have begun using the feature to generate weekly reports from raw CSV exports.
Step 1: Enable computer use in the Claude Console at console.anthropic.com and generate an API key. Step 2: Send a prompt that includes the current screenshot and a natural language instruction such as summarize last quarter sales. Step 3: Review the returned mouse and keyboard actions, approve them in the sandbox, and observe the model complete the workflow on screen.