2025.10.3

Release date: November 5, 2025

LLM configurations for Autopilot for Everyone

You can now configure your own model subscription for Autopilot for Everyone. This allows you to use models other than the default ones.

Read more about configuring LLM subscriptions.

Forcing unattended execution

Added the ability to force unattended execution with RunAsMe = false, allowing tools to run using the robot identity instead of the user identity.

Alternate startup agent

A new configuration option allows you to select an alternate startup agent, and choose between Specialized Autopilot or Conversational Agent.

Shortcut to tenant-level installation

Introduced a tenant-level Autopilot for Everyone card in the Admin section, accessible to tenant admins without requiring organization admin permissions.

Read more about the Autopilot for Everyone tenant card.

Conversational agent improvements

• Improved tool usage experience.
• Added support for file attachments.
• Improved tool use detail when running Agent as tool.

Misc improvements

• Improved consistency in time zone handling across Autopilot for Everyone components.
• The Autopilot for Everyone widget now requests focus only if the Autopilot for Everyone process is installed.

Bug fixes

• The specialized Autopilot was not able to use Context Grounding indexesc correctly.
• A mention user query issue affecting recognition in certain conversational contexts.
• In Autopilot for Everyone for Assistant Web, sessions could only restart after an idle timeout. Manual restarts now work as expected.

Documentation restructure

We’ve reorganized the Autopilot documentation to make it easier to navigate and to better reflect how Autopilot works across UiPath products.

The Autopilot for Everyone Admin Guide and User Guide are now part of the main Autopilot User Guide, enabling smoother navigation between general Autopilot information and Autopilot for Everyone–specific details.

The new Autopilot for Everyone chapter includes information about overview, installation, configuration, interaction, and plug-ins.

Autopilot remains the same AI-powered capability, but the experience differs depending on the product where you use it—for example, in Studio, Test Manager, or Assistant.

Additionally, we’ve transitioned from audience-centered to capability-centered documentation. The chapters previously titled Autopilot for Developers and Autopilot for Testers have been renamed to Generating automations and Generating tests, respectively.

This new structure helps you find the right guidance faster, organized by what you want to accomplish rather than by user type.