New

Admins can configure tool access permissions for Model Context Protocol (MCP) server integrations. When adding or editing an MCP server configuration, admins can select tools and set their access status as Allowed, Disallowed, or Require Approval. The interface supports bulk permission changes, search, authentication flows, and error handling for tool loading.
Admins can add Web applications as a communication channel under WebRTC. The WebRTC section supports adding new web applications, including an Engagement Messenger section with configuration links and read-only fields for Voice Service ID, Orchestrator Host URL, and WebSocket URL.
Admins can make retry attempts editable directly on the Caller Verification screen. The retry attempt value is now adjustable within the UI, improving workflow efficiency.
Caller verification supports a discard changes pop-up when refreshing configuration. Users are prompted to confirm before discarding unsaved changes during configuration refresh.
Caller verification includes a refresh configuration option. Admins can refresh configuration directly from the Assistant Designer page.
Caller verification now displays updated error messages for configuration issues. The UI provides clearer guidance for missing identification methods and factors.
Voice assistant handles form validation errors in Caller Verification. Mandatory fields not filled trigger a red card error, improving clarity.
Admins can select language input using DTMF or voice options. The Language and Voice page includes a new section for input selection, with links to advanced settings and retry/goodbye message configuration.
Admins can decouple the Testing UI from the Assistant "Active" state. Testing can be initiated without activating the assistant, and the test button is enabled regardless of activation status.
Admins can turn off parametric memory by default for all assistants. Scripts auto-populate parametric memory policy, ensuring consistent behavior.
Admins can now create a switch to enable custom SIP options in communication channels. The custom SIP option appears in the UI when enabled.
ServiceNow Otto panel premium chat is the default for eligible new customers, with Now LLM restrictions. Premium chat uses V2 orchestrator, and eligibility checks prevent Now LLM selection when premium chat and V2 orchestrator are enabled.

Changed

Embedded script for Service Portal UI has been updated. Parameter names, OAuth client ID handling, call button visibility, and optional widget parameters have been revised for improved reliability and clarity.

Fixed

Search sources in Assistant Designer now correctly reflect the "Include in AI Response" profile configuration. The system pulls from the correct table column, opens links in new tabs, and passes unit tests.
Unified Recommendations now allow cross-scope calls for typeahead suggestions, improving integration reliability.
The "Response feedback" page now works for existing assistants after upgrade. Script fixes ensure feedback attributes are correctly handled.
The "View org chart" feature in ServiceNow Otto for Virtual Agent now loads correctly when accessed through a custom URL or domain. The system uses relative URLs and sets appropriate roles.
Doc QnA now marks deflection logs as "response provided/resolved" after uploading an attachment with a question. The fix extends to all relevant topic representations and ensures logs are updated when user input is provided.
Duplicate create asset modal components have been eliminated from the codebase. Only one shared modal component is used across conversation-studio and nowassist-va repositories.

Removed

The role-based configuration from the MCP Server page is removed. The role column and related logic have been eliminated, simplifying server management and ensuring backward compatibility.
The Web applications section in Assistant Designer is no longer controlled by a system property. The property has been removed, and the section is always available.