UiPath ScreenPlay - Initiation

UiPath ScreenPlay can be installed in minutes and used either with UiPath’s built-in LLM subscriptions (OOB) or with your own AI subscriptions (BYOM), both in Studio and Studio Web. The following steps will walk you through everything a new user needs to go from zero to a working ScreenPlay automation.

 

Prerequisites in Studio / Studio Web

ScreenPlay works with any UiPath Studio version, but Studio 2025.10 or later is strongly recommended for a better prompt editing experience.​​

• For Studio (Desktop)

• Install UiPath Studio (preferably version 2025.10+).
• Ensure you are signed in to your Automation Cloud account from Studio (cloud.uipath.com).​
• Install the UiPath.UIAutomation.Activities package version 2025.10.20 or newer (or at least 25.10.5+ where ScreenPlay is available in preview).

• For Studio Web

• Access Studio Web from cloud.uipath.com, then go to Studio and create a new RPA Workflow.​​
• Make sure your Automation Ops policies allow preview activities and UIAutomation package versions that include ScreenPlay, if your organization uses governance.

 

Choose your deployment (OOB vs BYOM)

UiPath offers two main ways to power ScreenPlay. Your choice affects how you configure models, not how you build workflows.

• Out‑of‑the‑box (OOB)
• Uses UiPath-managed LLMs with ready-to-use models such as UiPath (with Gemini 2.5 Flash), UiPath (with GPT 4.1 / 4.1 mini), UiPath (with GPT 5 / 5 mini), OpenAI Operator, and Anthropic, Computer Use.
• No extra setup in AI Trust Layer is required for a basic start; you simply select a model in the ScreenPlay activity.
• Bring-your-own-model (BYOM)
• Lets you use your own subscriptions for Azure OpenAI, OpenAI, Amazon Bedrock (Anthropic Computer Use), Google Vertex, or any OpenAI V1‑compliant endpoint.
• Requires configuring an LLM configuration in AI Trust Layer so ScreenPlay uses your subscription instead of UiPath’s. 

 

Installing and using ScreenPlay in Studio

Once prerequisites are ready, ScreenPlay is added as an activity in your project.​​

• Create a project

• Open Studio and select a new Process (or other project template you prefer).​​
• Confirm you are connected to cloud.uipath.com (your Automation Cloud tenant).​

• Install UIAutomation with ScreenPlay

• Go to Manage Packages in the project.
• In Settings, ensure the Official feed is enabled under Default package sources.​
• Search for UiPath.UIAutomation.Activities and install a version that supports ScreenPlay (2025.10.20 or newer; 25.10.5+ for preview).

• ​​Add the ScreenPlay activity

• In the project canvas, click the Add Activity icon (or open Activities panel) and search for ScreenPlay.​​
• Drag the ScreenPlay activity into your workflow. It will typically appear inside a Use Application/Browser container after you indicate a target app window or browser.

 

• Configure your first prompt

• In the ScreenPlay Task field, write what you want ScreenPlay to do in natural language (for example, “Open Notepad, type today’s date, and save the file on the desktop”).
• From the Model dropdown, pick the default UiPath LLM or another available model (e.g., a GPT‑4.1 variant or Anthropic Computer Use) depending on what is enabled in your tenant.​​ 
• Adjust other properties (like timeout, maximum steps, or UI target hints) as available in your version.

 

• Run and refine

• Run the workflow from Studio and observe how ScreenPlay controls the application via mouse and keyboard interactions.​​
• If the result is inconsistent, refine the prompt, narrow the task, or break it into multiple steps using additional ScreenPlay activities.

​Installing and using ScreenPlay in Studio Web

In Studio Web the overall experience is similar, but accessed fully in the browser.​​

• Create a new RPA workflow

• Log in to https://cloud.uipath.com and open Studio.​
• Click Create New → RPA Workflow to create a new project.​

• Add the ScreenPlay activity

• Use the Add activity icon to search for ScreenPlay and add it to your workflow canvas.​​
• Ensure that preview activities are enabled via Automation Ops policies if ScreenPlay still appears as preview in your environment.

• ​Configure the activity

• In the Properties panel, provide the natural language prompt in the Task field.​​

 

• Select the Model from the dropdown (UiPath-provided model or one mapped via AI Trust Layer) and adjust any additional configuration fields.​​
•  

• Test from the browser

• Run the workflow directly in Studio Web; ScreenPlay will interact with indicated web or desktop targets, depending on how your environment is configured.​​

 Configuring BYOM in AI Trust Layer

If you want ScreenPlay to use your own LLM subscription, configure an LLM configuration in AI Trust Layer and map it to ScreenPlay.
​

• Open AI Trust Layer
• In Automation Cloud, go to Admin.​
• Select AI Trust Layer, then open the LLM configurations tab.​
• Create a ScreenPlay LLM configuration
• Select the appropriate Tenant from the dropdown, then click Add configuration.​
• In the Add configuration panel:
• Set Product to UI Automation.​
• Set Feature to ScreenPlay.​
• Set Configuration mode to Replace UiPath LLM subscription to tell UiPath to use your subscription instead of the built-in one.​

    Choose and connect your LLM

• From the LLM dropdown, select the LLM resource you want to use (e.g., Azure OpenAI model, OpenAI endpoint, Amazon Bedrock Anthropic, Google Vertex, or an OpenAI V1‑compliant endpoint configured via the OpenAI V1 Compliant LLM connector).
• In Configure customer managed connection, pick a Folder and then a Connector that matches your LLM provider (e.g., Azure OpenAI connector).
• Click Add new connection; a new tab opens where you enter keys, endpoints, and any other required information, then test and save the connection.

    Save and use in Studio

• Back in the Add configuration panel, click Test to validate the LLM configuration, then click Save.​
• After saving, your BYOM setup is available in Studio and Studio Web; ScreenPlay’s Model dropdown will expose that configuration according to your tenant and governance rules.

Best practices and precautions for first projects

• Because current LAM/LLM-based agents still have limitations, you get better reliability by designing simple, well-scoped ScreenPlay tasks.
• Use multiple, focused ScreenPlay steps
• Prefer several ScreenPlay activities, each responsible for one or two concrete actions on specific UI elements, instead of one long, end‑to‑end prompt trying to handle complex flows.
• Add guard-rails (like validation activities or checkpoints) between ScreenPlay steps to confirm that each step achieved the expected UI state.
• Be explicit in prompts
• Describe the UI context clearly (names of buttons, menu paths, specific text to look for) and what success looks like for the step.
• Include constraints in the prompt (for example, “Do not close the browser,” “Only modify the latest row,” or “Stop after the confirmation dialog appears”) to reduce unintended actions.
• Start in test environments
• Before pointing ScreenPlay at production apps, test workflows on demo or non‑critical environments to observe how the agent behaves in slightly changing UI conditions.