v26.10

link

UI Automation support for coding agents​

• Implemented semantic selectors for coding agents, enabling element identification using natural-language descriptions and visual analysis when DOM-based methods are unavailable.
• Added coding agent support for SAP GUI for Java applications.
• Added target configuration validation for coding agents, helping ensure targeting definitions are accurate before use.
• Updated fuzzy targets to disable check:text by default for Type Into, Set Text, Get Text, and Select Item activities, improving element targeting reliability.

• Expanded the uip rpa uia interact sap command set with expand-tree, select-dates-in-calendar, and click-picture-on-screen interactions.
• Added a new uip rpa uia target-anchorable preview-text command to preview the text a Get Text activity would return for a target anchorable definition.
• Enhanced the uip rpa uia target-anchorable resolve-defaults command to support element name and description attributes for more precise identification.
• Improved the command-line interface (CLI) experience on macOS with system permission inspection and a new uip rpa uia permissions command to verify or request necessary permissions (Accessibility and Screen recording).

UI Automation Modern​

• Adopted the new JS-enabled code editor widget for the Script Code property of Inject Js Script activity. The widget is available on Windows projects only, starting with Studio Desktop 26.0.198. Studio Web support is not yet implemented.
• Expanded UI Automation activity tracing to capture consumption details for Semantic Activities and ScreenPlay. Execution tracing is disabled by default and it can be enabled through the UI Automation Modern > Robot Logging > Enable activity tracing project setting. Trace data is uploaded to Orchestrator for failed jobs only.

Selection Helper UI Tree​

• Added an Enable Computer Vision button directly in the UI Tree panel. The CV elements tab is now always visible when an application is in scope and elements are being picked. When Computer Vision is disabled, the panel shows an actionable state with an Enable Computer Vision button that toggles CV on, captures a screenshot, and builds the CV element tree automatically.

Bug fixes​

• Fixed an issue where a proxy authentication dialog ("Sign in to access this site") would appear during element indication with Modern UI Automation activities, preventing users from indicating UI elements behind the dialog.

Selection Helper​

• Added support for configuring offset points directly from the Selection Helper, via a new button on the target expansion panel. When updating an offset point, the associated target is automatically selected and highlighted to provide immediate visual feedback.
• Updated the preferred UI framework selector to appear as a compact menu button in the action bar, with strategy abbreviations and icons (Auto/AA/UIA).
• Improved Selection Helper performance, reducing change-detection and rendering overhead during hover, element selection, and layout interactions.

UI Automation support for coding agents​

• Added Computer Vision (CV) targeting support, enabling coding agents to configure CV-based targeting methods for UI elements.
• Enhanced agent-based element discovery with image and semantic analysis capabilities. When a browser or application DOM is unavailable, agents can now identify UI elements using visual information and natural-language descriptions.
• Improved visual diagnostics by generating annotated screenshots that highlight identified elements, plus new screenshot-size caps to control image and token costs.
• Added support for validating target and anchor configurations through UI Automation tooling, helping ensure targeting definitions are accurate before use.

Improvements​

• Extended the UI Automation highlighting API with a Fill option that renders a solid-colored rectangle over a target element, enabling workflows to hide sensitive information or highlight multiple screen regions before capturing screenshots.

Agentic UIAutomation​

Semantic selectors​

Semantic activities​

ScreenPlay​

• SecureString and ScreenPlayImage support in coded workflows — ScreenPlay coded workflows now handle two additional argument types: SecureString, for passing sensitive string values without exposing them in plain text, and ScreenPlayImage, for including a screenshot directly in the ScreenPlay prompt to give the model additional visual context. Previously, coded workflows had no native handling for these types; this change brings them to parity with low-code workflows.
• Distributed tracing SpanId — ScreenPlay now adds an X-UIPATH-SEMANTIC-SPAN-ID header to the HTTP calls it makes to backend services, such as the Semantic Proxy and LLM Gateway. The span identifier lets a single ScreenPlay execution and its requests be correlated end-to-end across services, improving observability and troubleshooting.
• Application instance attach mode — ScreenPlay now supports the Application instance window attach mode on the Use Application/Browser activity. Inner activities search across the entire application instance — all parent and child windows, alerts, and popups — instead of a single window, and ScreenPlay captures and records traces correctly across all of them, in addition to the existing Single window support.
• Refreshed HTML trace file — The ScreenPlay HTML trace file has a refreshed visual design. The trace viewer output now has an updated layout and styling that make execution traces easier to read and navigate, aligned with current UiPath design standards.

Extract UI Data (Preview)​

Healing Agent​

UI Automation support for coding agents​

• Configure a target (uia-configure-target) — Point the agent at a screen and the elements you care about ("set up the Submit button on the checkout page"). It captures the live application, locates each element, generates a reliable selector, registers the screen and elements in the Object Repository, and verifies each one against a screenshot before handing back the reference IDs your workflow can use.
• Interact with a live app (uia-interact) — Ask the agent to interact with a running application — click, type, read values, fill forms, navigate menus, extract tables, or take screenshots. Useful for exploring an app before you automate it, or for confirming that a change behaves as expected.
• Improve a selector (uia-improve-selector) — Fix, harden, or recover a selector from runtime data. See Selector intelligence below.

Selector intelligence (uia-improve-selector)​

• Recover — the selector has stopped working ("element not found", a runtime failure). The agent rewrites it against the current state of the application so it matches the element again.
• Improve — the selector still works but is fragile, because it leans on position, an auto-generated ID, or a value that reflects changing content. The agent produces a more robust version that targets the same element.

• Driven by runtime data. It works from a live application or from data captured at the moment of failure — so a selector that broke on an unattended robot can be recovered later, on a different machine, even when the app is no longer running.
• Activity-aware. The agent knows, for example, that a Get Text target should not lean on the text it is meant to read, and that a Check target should not depend on whether the box is currently ticked. Each rewrite follows the rules that keep a selector stable for its specific use.
• Validated, not guessed. Every candidate selector is checked and scored before anything is written back. If the first attempt doesn't hold up, the agent refines it automatically and tries again, then keeps the strongest result.
• Explained. When a selector is recovered, you get a one-line root cause (why it broke) and the strategy used to make it durable — so you can trust the change at a glance.

Chromium Automation​

• Chromium Automation - Best for development, debugging, unattended automation with a visible browser
• Chromium Automation Headless - Best for unattended automation on servers and CI/CD pipelines

Oracle Automation​

macOS UI Automation​

• Enhanced support for macOS menu-based controls in Select Item activities. You can now use Select Item with AXPopUpButton and AXMenuButton controls, enabling reliable interaction with dropdown and menu-based UI elements that require opening a popup to access their items.
• Added support for triple-click actions. This capability is available on:
  • Windows and macOS using Hardware Events
  • Windows and macOS using ChromiumAPI
• Added support for mouse drag actions on macOS.

UI Automation Modern​

• Enabled the Capture Elements feature in Object Repository in cross-platform projects.

• The Set Runtime Browser activity can now resolve the default system browser at runtime in cross-platform projects. A new Use default system browser option allows selecting between the system default browser or a specific browser:

  • Windows projects: Chrome, Edge, Firefox
  • Cross-platform projects: Chrome, Edge, Safari

  The activity’s display name now reflects the selected browser (literal value, Default system browser, or a truncated expression).

• Added the Accessibility Check activity which enables users to perform Accessibility Testing in Test Manager instances within Test Cloud. This activity loads a web browser page, and performs accessibility checks as per the WCAG 2.1 (or higher) standard. Examples of accessibility checks include DOM and visual accessibility, contrast ratio, alt tags, tab order, improper ARIA roles.

Advanced Selector Editor​

• Added zoom controls to the Advanced Selector Editor and Selection Helper, allowing you to zoom in or out between 80% and 110% using +, -, and Reset options.

• Added case-sensitive support for attributes in the Advanced Selector Editor.

• Added an Edit target option to Use Application/Browser, exposing:

  • Window selector
  • Application properties (such as Browser URL, browser arguments, application path, application arguments, and working directory)
  • The option to Update the window selector in child activities. This allows any changes made to the screen’s Window Selector to be automatically applied to all child elements.

• The Advanced Selector Editor can now be used for editing Window selectors in addition to UI element Selectors.

• In the UI Tree panel of UI Explorer, added a toggle to show or hide properties with empty values in the Selector compatible and Other(s) sections, helping display only relevant information.

• Added a Filter properties option to the UI Tree panel, enabling filtering of properties for the selected node by name in both the Selector compatible and Other(s) sections.

Improvements​

• Optimized Active Accessibility selectors to improve Extract Table Data performance for large tables.

Bug fixes​

• For link type nodes, the text was not retrieved and the tree path was generated using the internal SAP id. With this fix, the proper node text is used. This affected the SAP Expand Tree activity.

• Fixed an issue where selectors could not be identified when interacting with HTML views in the CAMBIO Cosmic application (a Java-based healthcare application using JxBrowser).

• Fixed an issue where the cross-platform Type Into activity became unresponsive on Automation Cloud Robot VMs.

• Fixed native capture for glyphs and Japanese characters.

• Fixed an issue where Java extensions continued to be reported as installed after uninstalling them.

Corrections, added August 11, 2026​

Expanded model selection for ScreenPlay​

• Screen Agent (using Path Mini)
• Screen Agent (using Gemini 2.5 Flash)
• Screen Agent (using Gemini 3.0 Flash Preview)
• Screen Agent (using GPT-4.1 mini)
• Screen Agent (using GPT-5 mini)
• Screen Agent (using GPT-5.4 mini)

• Screen Agent (using GPT-4.1)
• Screen Agent (using GPT-5)
• Screen Agent (using GPT-5.4)
• Screen Agent (using GPT-5.5)
• OpenAI - Operator
• Anthropic - Computer Use

New default model for ScreenPlay​

ScreenPlay execution beyond the indicated scope​

ScreenPlay execution fixes​