Cloudera Docs

Agent Studios - TP (version 2.1.0)

Release notes and fixed issues for version Agent Studio 2.1.0.

New Features / Improvements

  • Artifact Support: Agents can now reliably create, read, and share files such as CSV, JSON, images, and custom code within workflows. This allows agents to handle and exchange large datasets and rich outputs effectively.
  • Smart Workflows: When enabled, the Smart Workflows feature uses an upgraded agentic framework that improves context management in complex multi-agent workflows, particularly those orchestrated by a Manager Agent. It ensures that agents receive the necessary historical context, reducing hallucination rates and improving overall workflow accuracy and stability.
  • Planning and Thought Bubble: When enabled, the Manager Agent can now generate an explicit, multiple-step plan to dynamically decompose complex user queries into smaller, manageable subtasks. The Thought Bubble feature provides real-time visibility into the decision-making process, task delegation, and planning status of the agent.
  • Tools Playground: Tools Playground is a dedicated UI now available for developers to test and validate custom tools and their parameters, including Artifact handling, in isolation before integrating them into a live workflow.

For more information on using Artifact Support, Smart Workflows, Planning and Thought Bubble, and Tools Playground, see Agentic Workflows .

  • Secure Tool Execution and Development: The Secure Tool Execution and Development feature is a security enhancement ensuring that all tool runs are performed within a secure, sandbox environment. This isolation prevents malicious or unintended code execution from compromising the host system or other processes.

    For more information, see Secure Tool Execution and Development .

  • Agent Studio installation using ML Runtime image: Agent Studio is now exclusively shipped as a prebuilt, containerized ML Runtime Image for both on-premise and cloud deployments. This streamlined approach eliminates the need to compile source code within the customer environment, providing a consistent, reliable, and production-ready installation.
    • Migration Support: Migration to ML Runtime Image mode is supported for the existing Agent Studio deployments.
    • Deprecation Notice: The existing methods of deployment (from source code) and AMP mode are now deprecated. The latest releases will be available only through an ML Runtime Image.

    For more information on how to install, see Deploying Agent Studio using the ML Runtime Image .

  • Service Account support: Agent Studio supports deploying production and shared workflows using a Service Account (Machine User) identity and its corresponding namespace. This provides enhanced security and fine-grained permissions compared to deploying using a human account.

Fixed Issues

  • A critical remote code execution vulnerability (NVD - CVE-2025-55182) in React Server Components has been fixed. For more information, see TSB 2025-921.
  • Fixed an issue that caused API keys to be logged in plain text within Phoenix, addressing a significant security vulnerability. (DSE-48556)
  • Resolved issues preventing the suspend and resume functionality of workflows from working correctly. This fix also resolves related UI inconsistencies. (DSE-48347)
  • Fixed Model Context Protocol (MCP) server registration and configuration issues to ensure that third-party tools integrate reliably. (DSE-49771)
  • Addressed MCP server functionality and stability issues impacting tool runs within workflows. (DSE-49772)
  • Updated and refined several Workflow Templates to align with new features and improve the starting point quality. (DSE-49783)