Anthropic tests “Tasks Mode” for Claude, aiming to build a fully autonomous AI Agent

Anthropic is currently testing a new “Tasks Mode” for its AI assistant Claude, designed to transform it into a powerful agent capable of managing complex, multi-step workflows.

In the latest internal build, Claude features an intuitive dual-panel interface:

• The right panel displays user-uploaded files or AI-generated context.
• The left panel shows a live, step-by-step execution timeline, making task progress transparent and logic interpretable.

Unlike conventional AI assistants, Claude first generates a structured “Action Plan” when given a complex request. If instructions are ambiguous, it proactively asks clarifying questions—and if no response arrives within a set time, it intelligently skips ahead to keep the workflow moving.

Users can intervene at any time to adjust the plan or add new context, ensuring outputs align with their intent. Claude dynamically adapts in real time, demonstrating high flexibility and responsiveness.

The mode also integrates Skills and Model Context Protocols (MCPs), enabling automatic tool use—such as code generation, web search, or API calls—based on task requirements. All outputs, including code, documents, or web apps, are automatically saved under a dedicated “Artifacts” tab for easy preview and review, forming a complete, traceable workflow.

Industry observers note that the update appears aimed at attracting users of Google’s NotebookLM, though Claude’s approach offers deeper task planning and multi-artifact support. While Anthropic hasn’t confirmed a public launch date, the feature’s maturity suggests a rollout could come as early as this month.