July '26 Release Notes

New Features and Models

• New Login Flow with Sign-up Restriction: The sign-in experience now automatically remembers your email and authentication provider, eliminating the need to display all login options. Additionally, an organization can restrict the sign-up of new users based on email domain, restricting users in their organization from registering self-service.
• Studio Assistant: The Studio Assistant is a powerful agentic assistant embedded directly inside Vertesia Studio. We created a specialized built-in agent equipped with new skills and tools, enabling the Studio Assistant to automate everything that you previously did manually in Studio. Create, test, and optimize agents with natural language instructions. You can also use the Studio Assistant to create and manage content, collections, types, databases, dashboards, and even our new workflow processes. 
• Process Engine: We've created a new process engine for defining, executing, and monitoring multi-step automated workflows. Use the Studio Assistant to autonomously configure, test and publish new processes based on an image, document, technical configuration (like BPEL), or a conversation about the workflow process. The accompanying "Task Inbox" lets you review and act on pending tasks generated by processes.
• Agent Scheduling: Configure any agent to run on a recurring schedule (daily, weekly, monthly, or any custom interval). These agents run like any other agent with your identity, permissions and tool access, but on a schedule.
• Browser Use Agent Tooling: Agents can now navigate and interact with websites as part of automated workflows using built-in Browser Use capabilities configurable at the project level. This enables web-based research, form filling, and scraping tasks to be handled directly by agents without custom integrations.
• Vertesia Remote MCP: Connect any coding agent to Vertesia's MCP and create your app along with all of your Studio configuration, directly from the coding agent. This new remote MCP exposes the same skills and tools behind the Studio Assistant, but for those working outside of Vertesia’s AI Studio.
• Gemini Embedding 2: Vertesia now supports the gemini-embedding-2 model, and is used as the default embedding model with Google Agent Platform (VertexAI) environments.

Improvements

• Rebooted Agent Conversation and Workstream Chat UX: The agent conversation has been redesigned with a cleaner workstream chat view and better handling of multi-turn interactions, inspired by the most popular agent experiences.
• Interactive Agent Approval Modes: Agents now support configurable approval modes that require human review before proceeding with certain actions, optionally inserting human-in-the-loop approval requests.
• RTL Language Support: Vertesia now supports right-to-left (RTL) languages with locale-aware date and number formatting, making Studio accessible to users working in RTL language environments.
• WCAG 2.1 AA Accessibility: Vertesia Studio and plugin templates now meet WCAG 2.1 AA accessibility standards, improving the experience for users who rely on assistive technologies.
• Overflow Tabs Into "More" Dropdown: When tabs exceed the available screen space in the agent right panel or across workstream tabs, the extra tabs automatically collapse into a "More" dropdown menu, ensuring all panels remain visible without scrolling.
• Configurable Header Menu for Composite Apps: Composite app administrators can now fully configure the top header bar, building an ordered list of buttons that combines built-in navigation controls along with custom admin-defined buttons.
• Interaction Deletion with Cascade Option: Deleting an interaction now supports a cascade option that removes dependent objects in a single operation, reducing the manual cleanup associated with interaction hierarchies and previously published versions.
• Auditor and Support Read-Only Roles: Two new read-only roles — Auditor and Support — are now available in Studio for compliance monitoring and support workflows without write access. Auditors can review all platform activity; Support users can access user-facing views to assist customers.

Bug Fixes

• Runs and Agent Runs Tables: The "Runs" and "Agents Runs" tables have been updated to avoid any co-mingling of direct LLM calls and multi-turn agents.
• Sub-Agents Use Their Configured Tool Set: Sub-agents correctly use the configured list of tools and skills.
• Prompt Rendering Consistency: Prompt rendering correctly handles content that mixes plain text and template expressions, preventing rendering inconsistencies when switching between modes.
• Skill Tools Return Interaction Prompt: Interactions configured as a skill tool return the interaction prompt when invoked, instead of the interaction description.
• MCP OAuth Discovery for Split-Domain Providers: MCP OAuth discovery, authentication and tool access completes successfully for providers using separate domains for MCP and API endpoints.
• Infinite Scroll on Content Lists: A regression caused the loading spinner to appear without fetching additional items; scrolling to the bottom now correctly loads the next page.
• Bulk Content-Object Updates with String Type: Bulk updates to content objects that include a string-type field are correctly persisted.
• Enterprise SCIM/UPN Users Unable to Sign In: Authentication allows SCIM users to login with either an email or username.
• Plugin CSS Isolation: Plugin CSS styles are automatically repaired in unstyled Shadow DOM by setting an optional ui.css_rebuild property to true in the application manifest.

Deprecated

• Custom Workers: The new standard workflow for bulk importing content has replaced custom workers used for bulk import.
• Legacy CLI Commands: The worker, memory, codgen, and datagen legacy commands have been removed from the CLI.