🧭 Claude.ai — Chained Task Automations for No-Code Multi-Step Workflows

Anthropic has introduced chained task automations in Claude.ai, a new feature that allows users to define sequences of connected tasks that Claude executes in order without requiring the user to stay present for each step. The feature is available on Pro, Max, Teams, and Enterprise plans, and is designed for the recurring workflows that users currently manage manually — research pipelines, document formatting sequences, and data extraction chains that involve multiple distinct processing steps.

How chained automations work

• Visual builder — a drag-and-drop interface in the Projects section of Claude.ai allows users to define automation steps, connect outputs to subsequent step inputs, and set conditions for branching
• Trigger options — automations can be triggered manually, on a schedule, or when a new file is added to a linked Project document store
• Intermediate review points — users can insert review checkpoints between steps where Claude will pause and surface the intermediate result for approval before proceeding
• Run history — a log of past automation runs with step-by-step output visibility, enabling debugging of failed or unexpected runs

Anthropic notes this is an early version of automation capabilities and that the feature is designed to complement rather than replace the Claude Code CLI for developers who need more programmatic control.

🧭 Computer Use — Multi-Screen Desktop Automation Now Supported

Anthropic has updated the computer use API capability to support multi-screen desktop environments. The initial release of computer use operated on a single virtual display; the updated implementation allows Claude to navigate and interact across multiple monitors when they are configured as part of the virtual desktop environment provided to the computer use session. This is a practical requirement for enterprise automation workflows where target applications span multiple screens — financial dashboards, engineering toolchains, and data analysis environments that commonly use two or three displays.

Technical changes

• Multi-display screenshot — the screenshot tool now accepts an optional display_id parameter; omitting it captures the primary display, while passing specific IDs captures any configured display
• Coordinate system — click and mouse-move coordinates are now expressed in the display-space coordinate system of the specified display, avoiding the coordinate translation errors that occurred when multi-screen environments were approximated as a single large canvas
• Context efficiency — Claude now requests screenshots of specific displays rather than the full composite environment, reducing image tokens consumed per action by approximately 40% in two-monitor setups