> ## Content Index
> Fetch the complete content index at: https://www.testingcatalog.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Exclusive: Early look at Claude Tasks Mode from Anthropic
- URL: https://www.testingcatalog.com/exclusive-early-look-at-claude-tasks-mode-agent-from-anthropic/
- Published: 2025-12-18T14:04:40.000Z
- Updated: 2026-01-19T10:32:17.000Z
- Description: Anthropic’s Claude task mode will offer step-by-step task execution, artefact management, and context file control in a streamlined dual-panel UI
- Author: Alexey Shabanov
- Tags: Claude News, AI Rumours, Latest AI News

Anthropic’s new [task mode](https://www.testingcatalog.com/anthropic-testing-new-agentic-tasks-mode-for-claude/) for Claude is now entering a functional phase, with internal users able to send tasks directly to the model and observe its step-by-step execution. The updated UI presents a dual-panel layout: uploaded or model-generated context files are visible in a right-side sheet, while a new progress section dynamically tracks the execution of multi-step tasks. 

> BREAKING 🚨: Early look at Claude Task Mode agent workflow. What do we know so far?  
>  
> \- Claude Tasks will operate various Skills and MCPs to achieve the goal.  
> \- Claude will ask clarifying questions before the execution or skip it after a timeout.   
> \- Claude will generate an action… [pic.twitter.com/JvKN9G8H2T](https://t.co/JvKN9G8H2T?ref=testingcatalog.com)
> 
> — TestingCatalog News 🗞 (@testingcatalog) [December 18, 2025](https://twitter.com/testingcatalog/status/2001654962753851721?ref%5Fsrc=twsrc%5Etfw&ref=testingcatalog.com)

When complex requests are submitted, Claude generates a structured plan, displaying each step as it progresses. This setup enables users to review, edit, or select specific context elements or “[skills](https://www.testingcatalog.com/anthropic-prepares-clause-skills-update-and-a-new-model/)” the agent utilizes during task completion, further personalizing the workflow.

![Claude](https://storage.ghost.io/c/2a/1b/2a1b1782-8506-4d7d-bf53-ad3fb52e2a0f/content/images/2025/12/Claude-features-showcase-app-Claude-12-18-2025_11_49_AM.jpg)

Claude Task Mode Agent UI

The system can prompt users to make selections or input custom details when needed. Once an artifact, such as code, an article, or a web app, is generated, it appears in a dedicated artifact tab for immediate review. 

![Claude](https://storage.ghost.io/c/2a/1b/2a1b1782-8506-4d7d-bf53-ad3fb52e2a0f/content/images/2025/12/Claude-features-showcase-app-Claude-12-18-2025_11_42_AM.jpg)

Claude will ask clarifying questions before executing the task. This step will be skipped automatically if there is no response. 

This approach draws parallels with NotebookLM’s interface, which also relies on panels for context management and output navigation. However, Anthropic’s implementation introduces deeper task planning and execution, likely influenced by its work on Claude Code.

![Claude](https://storage.ghost.io/c/2a/1b/2a1b1782-8506-4d7d-bf53-ad3fb52e2a0f/content/images/2025/12/TestingCatalog-News-Claude-12-18-2025_11_10_AM.jpg)

Claude will operate different Skills and MCPs in order to achieve the goal. For complex tasks it will generate an action plan that needs to be executed.

These changes position Claude to appeal to users currently relying on tools like [NotebookLM](https://www.testingcatalog.com/tag/notebooklm/), aiming to capture scenarios that demand robust context transformation and multi-artifact output. 

![Claude](https://storage.ghost.io/c/2a/1b/2a1b1782-8506-4d7d-bf53-ad3fb52e2a0f/content/images/2025/12/TestingCatalog-News-Claude-12-18-2025_11_12_AM.jpg)

Users will be able to modify Claude's action plan on the fly during the execution and send additional context to Claude Tasks workflow. 

By integrating planning and artifact management into a cohesive workflow, Anthropic demonstrates its strategy to position [Claude](https://www.testingcatalog.com/tag/claude/) as a comprehensive agent for knowledge work, code generation, and complex task automation. While the official release timeline is not yet confirmed, the rapid pace of visible updates suggests this feature could become available in the near term, possibly within December, depending on final development and testing progress.