What's new in IBM® watsonx Orchestrate®

Support for FedRAMP through AWS GovCloud (US)

IBM watsonx Orchestrate is now available on AWS GovCloud (US), a secure and cloud-based environment designed to support U.S. government agencies and organizations handling sensitive and regulated data. It provides a compliant infrastructure to meet stringent federal requirements. AWS GovCloud (US) inherits the robust compliance framework of AWS, enabling customers to align with critical standards such as:

• Enhanced compliance alignment: AWS GovCloud (US) now offers improved support for FedRAMP requirements, streamlining authorization processes for government workloads.
• Strengthened data sovereignty: Expanded operational controls ensure that all regions remain managed exclusively by U.S. citizens, reinforcing compliance with government regulations.
• Improved isolation architecture: Logical and physical isolation enhancements provide greater security for sensitive workloads.

For more information, see AWS GovCloud (US): compliance overview.

Import external A2A agents as collaborators

You can now add external agents that use the Agent to Agent (A2A) protocol as collaborators when you build an agent in the UI. This update makes it easier to integrate third-party systems that follow the A2A standard and control how they exchange data during execution. For configuration details, see Adding agents from third-party platforms.

Data isolation support for AWS HIPAA-regulated environments

Deploy watsonx Orchestrate on AWS HIPAA-regulated clusters with data isolation. This capability provides dedicated data storage for stricter security requirements. For more details, see Data isolation on watsonx Orchestrate.

View tool runtime calls in observability traces

You can now see tool execution details inline in the Trace Details view when a flow includes a tool call. This update gives you full visibility into agent behavior and helps you troubleshoot faster. For more information, see Debug tool flow runtime issues.

Support single step SSO for embedded chat

You can now use single-step SSO token exchange for embedded chat. This update introduces a new authentication type in the UI that simplifies setup and improves SSO compatibility. Use this option if your identity provider supports direct token exchange for access tokens. For more information, see Configuring single sign-on for applications.

New AI model available for agent creation

You can now use GPT-OSS 120B — OpenAI (via Groq) when creating or editing an agent. This model is available in the AI model drop-down and can be used for supported use cases that require ultra-low latency or cost-efficient, high-volume processing.

Add a phone-like preview for testing DTMF in voice agents

You can now use phone-like preview in voice chats to test DTMF settings without ending the voice chat or making an actual phone call. By using the keypad, you can send DTMF inputs (0–9, *, #) to the agent and validate DTMF-dependent flows. You can also choose to turn off the setting without ending the ongoing voice session. For more information, see Configuring voice settings for agents.

Add audio cues for long voice agent responses

Voice agents can now play built-in “typing” sounds and hold music when a response takes longer than expected. These audio cues reassure callers that the agent is still processing their request and help reduce confusion during long pauses. For more information, see Configuring voice settings for agents.

Add DTMF support for SIP and Genesys audio connector

You can now enable Dual-Tone Multi-Frequency (DTMF) input for phone calls that use SIP or Genesys Audio Connector. By using DTMF, your voice agents can collect keypad inputs during a call, supporting IVR-style flows and mixed voice–keypad interactions. You can configure DTMF settings in Manage > Voice when you create or edit a voice configuration. For more information, see Configuring voice settings for agents.

Add Deepgram as a Speech to Text and Text to Speech provider

You can now select Deepgram as a Speech to Text (STT) and Text to Speech (TTS) provider when you create or edit a voice configuration. This update gives you additional language options and faster response times for voice interactions. To use Deepgram, go to Manage > Voice, create or edit a voice configuration, and choose Deepgram in the STT or TTS provider list. The interface shows only the settings that are supported for the selected Deepgram model and language. For more information, see Configuring voice settings for agents.

Enhanced document processing for knowledge sources

When you upload documents to a knowledge source, they are now processed by using advanced extraction technology for higher accuracy and richer content quality.

Key features:

• Preserves the structure of tables for more accurate representation.
• Extracts text from images ensuring that no critical information is missed.
• Stores extracted content in Markdown to improve readability and context.

For more information, see Enhanced document processing capabilities.

Copy agent responses

You can now copy any agent response, including plain text and grid responses, directly from Orchestrate Chat. Select the Copy icon next to a response to copy the full output with its formatting, making it easy to reuse content in documents, emails, and workflows.

New prebuilt and partner agents available in the catalog

Other enhancements and changes

The following table lists the enhancements available in the current release corresponding to the user role:

Other enhancements and changes

Documentation updates

See the updates in the watsonx Orchestrate documentation and other contents for this release.

New sections

• The Available AI models section lists all models that you can use in watsonx Orchestrate.

• The Adding AI models through AI model gateway section explains how to integrate models from different providers by using AI Gateway.

• The Design guidelines for agent icons section describes the specifications that you must follow when you upload an agent icon.

Updated sections

• The Connecting to an Astra DB content repository, Connecting to an Elasticsearch content repository and Connecting to a Milvus content repository sections now include guidance for passing context variables.

• The Choosing an AI model section now explains how to choose the right model for your use case.

• The Considerations for choosing an AI model section now provides more detailed guidance to help you evaluate model requirements.

• The Configuring voice settings for agents section now includes support for:

  • Phone-like preview for testing DTMF in voice agents
  • Audio cues for long agent responses
  • DTMF support for SIP and Genesys Audio Connector
  • Deepgram as a Speech to Text and Text to Speech provider

• The Adding agents for orchestration section was fully reorganized to improve clarity and navigation. The main page is streamlined, and new subpages group the different ways you can add agents, including:

  • Adding agents from catalog
  • Adding an agent from a local instance
  • Importing agents from external platforms
  • Adding agents from third-party platforms
  • Adding watsonx Orchestrate assistants as collaborator agents
  • Adding AI assistants as collaborator agents

• The Known issues and limitations topic now has the following entry:

  Chat

  • Orchestrate chat scrolling in Safari

Simplified navigation for building and analyzing agents

watsonx Orchestrate now features a redesigned navigation to help you build and analyze agents more efficiently.

• Analyze is now available directly in the main menu.

  Use this option to quickly access the Agent analytics page. Agent analytics in watsonx Orchestrate provides built-in monitoring and observability, enabling builders to track, analyze, and optimize the performance of agents.

• Assistant builder is moved to the bottom of the main menu to help you focus on creating and managing agents and tools.

• The Build menu now opens the Build agents and tools page directly for quicker access.

This update gives you a clearer, more organized workspace. For more information, see Building agents.

Manage all your knowledge sources from one place

You can now manage all your knowledge sources in a single, centralized location with the new All knowledge section in watsonx Orchestrate.

• You can create a knowledge source independently of any agent from the All knowledge page.
• You can create a knowledge source directly from the Agent Builder page, and it is automatically added to your agent after the creation.
• You can add an existing knowledge source to an agent, allowing you to reuse a single knowledge across multiple agents.
• Each agent continues to support one knowledge at a time, but a knowledge can now be shared across agents for consistency, and easier maintenance.

This improvement streamlines how builders manage and reuse knowledge sources, reducing setup duplication and simplifying updates in agents. For more information, see Building a knowledge source.

Purchase partner agents from the catalog on IBM Cloud

You can now purchase and use partner agents directly from the watsonx Orchestrate catalog on IBM Cloud. This in-app experience enables you to discover, review, and subscribe to partner-built agents without leaving the watsonx Orchestrate workspace.

Partner agents are prebuilt AI agents developed by IBM-approved external partners and available in the watsonx Orchestrate catalog. Agents that require a subscription are marked with a lock icon and can be unlocked after purchase. Once purchased, the agent becomes available for builders in your instance. Builders can start using the purchased agent as a template to create a new agent or add it as a collaborator agent. This enhancement simplifies the way admins and builders discover, subscribe to, and manage partner-built agents, providing faster access to ready-to-use automations. For more information, see Purchasing partner agents.

The following GIF illustrates how admins can purchase partner agents from the catalog.

Note: This video has no narration.

Handle changes in subscription ownership more seamlessly

You can now manage changes in your subscription ownership or admin roles more smoothly on watsonx Orchestrate.

When a new subscription owner or admin is assigned in the SaaS Console, watsonx Orchestrate now reflects the changes consistently across the subscription and its instances. Currently, when new admins are added, the changes do not appear in the MCSP SaaS Console. Only subscription owner changes are synchronized with the MCSP SaaS Console. This update helps ensure that all subscription admins have access to existing and newly created instances without disruption.

Support for latest A2A protocol version

watsonx Orchestrate now supports version 0.3.0 of the Agent-to-Agent (A2A) protocol for external A2A agents. This update ensures compatibility with the latest protocol capabilities and improves reliability when integrating external agents.

Easily connect to MCP servers

You can now easily locate where to add MCP servers when building agents. The Add a tool modal is redesigned to make it simpler to discover and connect to MCP servers, OpenAPI specs, or tools from local instance. This enhancement improves discoverability and usability for builders who need to add or manage MCP tools efficiently. For more information, see Importing tools from an MCP server.

Data center expansion in IBM Cloud

The IBM Cloud data center is now expanded to include the following region:

São Paulo

IBM watsonx Orchestrate data center now includes São Paulo. This expansion helps improve performance, reduce latency, and meet data sovereignty requirements for users in Latin America. For more information, see Regional availability and outbound IP addresses.

Newly added prebuilt agents and tools

IBM watsonx Orchestrate introduces new prebuilt agents and tools to expand automation across key business domains.

For the complete list of available prebuilt agents and tools, see List of prebuilt agents.

IBM® watsonx Orchestrate® API updates

See the updates in the IBM watsonx Orchestrate API for this release.

Other enhancements and changes

The following table details the fixes that are delivered in the current release corresponding to the user role:

Documentation updates

See the updates in the watsonx Orchestrate documentation and other contents for this release.

New sections

• The Connecting to GitLab section shows how to enable the integration with watsonx Orchestrate.

• The Collecting diagnostic data for performance issues section provides instructions for gathering data for performance-related support cases.

• The Building a knowledge base section provides guidance on adding knowledge to agents.

Updated sections

• The Embedding agents in applications section was updated to add information about the noneditable agents.

• The Connecting agents to channels and Connecting agents to chat sections were updated to add information about voice interaction on the Orchestrate Chat home page.

• The Connecting to IBM Planning Analytics section was updated to add information about API key requirement in Basic Authentication.

• The Configuring the agent's behavior section was reorganized and improved to maintain consistency.

• The Connecting to Slack was updated to include steps to enable email visibility on Slack.

• The Known issues and limitations topic now has the following entry:

  AI assistant access failure

  • AI Assistant access failure for admin users added through SSM

  IBM SaaS Console

  • Subscription access in SaaS Console

  Agents

  • Purchasing partner agents