Agents release notes

We’re excited to introduce conversational agents, a new agent type designed for dynamic, multi-turn interactions powered by natural language. Built directly in Studio Web, conversational agents follow the same low-code design experience as regular agents, making it easy to define system prompts, configure tools, connect to context, and add human escalations.

These agents are ideal for use cases like self-service support, guided workflows, and interactive Q&A. They understand context across turns, respond to user input in real time, and can trigger actions or escalate when needed. Built-in testing and evaluation tools allow you to validate behavior before deployment.

For details, refer to Conversational agents.

You can now better manage traces visibility using a combination of the existing Logs.View and Jobs.View permissions. This update introduces fine-grained access controls that lets admins tailor what trace data users can see, addressing security and compliance needs. Depending on permissions, users may view all attributes, a limited subset, or no trace data at all.

For details, refer to Traces.