About Autopilot Chat

What is Autopilot Chat?

Autopilot Chat is a context-aware and action-enabled agent which is available in Test Manager in Public Preview. Autopilot Chat brings an intelligent, conversational assistant in Test Manager, designed to enhance how testers interact with and manage their assets.

Instead of navigating menus and filters, users can ask questions, perform searches, generate assets, and trigger actions in natural language. The chat maintains conversational context when using tools, making interactions more natural and fluid.

Autopilot Chat connects with Test Manager’s built-in Autopilot tools and extends further through MCP server integration, allowing it to execute commands, evaluate requirements, generate test cases, and even run external automation processes and agents.

Important:

Autopilot Chat is available in Test Manager only when delivered through Test Cloud.

Figure 1. Autopilot Chat - generic window

For additional information and recorded tutorials, subscribe to the complementary information maintained on the UiPath Insider Portal where you can also leave feedback.

How many modes of working does Autopilot Chat have?

Autopilot Chat can be used with 2 modes: Edit mode and Agent mode.

1. Edit mode enables users to query the system and then choose one of the tools or operations suggested by Autopilot to perform the queried task. Autopilot does not proceed unless users confirm the tool to be used.

Figure 2. Autopilot Chat - Edit mode

When working in Edit mode, you can consult the reasoning and expand it.

Figure 3. Autopilot Chat - Edit mode reasoning

2. Agent mode enables users to query the system but then fully rely on Autopilot to choose the best tools and operations to perform the queried task (with no human intervention). Autopilot proceeds automatically and no user confirmation is required.

Figure 4. Autopilot Chat - Agent mode

What LLMs can be used with Autopilot Chat?

The list of supported LLMs can be found in the Configuring LLMs topic.

How can I access Autopilot Chat from Test Manager?

To access Autopilot Chat, look for the Autopilot icon in the global header. Open a chat, access the Settings menu or the Chat History. To learn more about the chat screen and the available settings, refer to Autopilot - User interface.

What functionalities does Autopilot Chat cover in Test Manager?

Instead of navigating menus and filters, users can ask questions, perform searches, generate assets, and trigger actions in natural language.

Open the Autopilot Chat settings.

Figure 5. Autopilot Chat settings

The Test Manager framework tools are:

• Document Search
• Autopilot Search
• Web Search
• Requirement Evaluation
• Test Generation
• Obsolete Test Detection
• Autopilot Chart

Figure 6. Test Manager framework tools

The Test Manager MCP tools are:

• Assign test cases to requirements
• Assign test cases to test set
• Assign object label
• Bulk create test cases
• Create requirement
• Create test case
• Create test set
• Get all labels
• Get all object labels
• Get requirements
• Get test cases
• Get test sets
• Get test executions
• Get test case logs for test executions
• Unassign test cases from requirement
• Unassign test cases from test set

Figure 7. Test Manager MCP tools

To learn more about Autopilot Chat across the UiPath suite of products, refer to Autopilot - About Autopilot chat.

Can I connect a custom MCP server to Autopilot Chat in Test Manager?

Only the custom MCP servers created in Orchestrator are supported. Open the Autopilot Chat settings and configure the MCP servers as described in the Autopilot - User interface topic.

Are there any query and prompt guidelines for Autopilot Chat?

To learn more about query and prompt guidelines, refer to Autopilot - Interacting with Autopilot chat - recommendations.

Is there support for queries which contain custom fields?

Is there support for queries which contain custom fields? Yes.

The queries support both text-based custom fields and label custom fields. Use Autopilot Chat for custom field operations such as:

• Find all the available custom fields (example prompt: Show me all the custom fields).
• Find data objects which bear one or more custom fields (example prompt: Find all requirements where Sprint = S1 and owner = John Smith).
• Move data objects (example prompt: Move requirement REQ:2 from S3 to S4.)

What are the regions Autopilot Chat is deployed in?

Autopilot Chat is deployed in the following regions: EU, US, JA, CA, AUS, SI, UK, IN. For more information, refer to the Global cloud regions section of the documentation.

The LLMs which support Autopilot Chat are deployed in the following regions: EU, US, JP. For all the other available regions, the LLMs are re-routed. For more information, refer to the AI features and model routing section of the documentation.

Prerequisites for Autopilot Chat

1. Make sure you have a UiPath Test Cloud account.
2. Make sure to enable AI models (LLMs) in your AI Trust Layer policies. To use the AI capabilities provided by Autopilot, make sure you have an AI Trust Layer policy active and deployed to your tenant. Learn how to:
   • Create a policy
   • Configure the policy for AI Trust Layer
   • Deploy the policy to the tenant where you want use Autopilot
3. Make sure you have prepared any documents you want to attach as part of your query. Autopilot Chat accepts PDF, TXT, MD, Word, and Excel formats. Note that, if you attach PDF documents, Autopilot Chat takes into account its structure and formatting during processing.
4. Before entering your query into the Autopilot Chat window, make sure that you choose the working mode (Edit mode or Agent mode).

Figure 8. Autopilot Chat mode

Functionalities in Autopilot Chat

Autopilot Search

Use Autopilot Search directly inside the chat, by typing full, partial or fuzzy terms of what you are looking for. The agent retrieves all the matching objects. If a typo is made, the chat auto-corrects common mistakes.

After results appear, you can expand them or open the filtered view in the artifact table, without leaving the chat.

Query examples:

• Find all requirements where custom field Sprint is set to 123.
• Find all the test cases that failed in the last 5 days.

Figure 9. Autopilot Search in Autopilot Chat

Autopilot Charts

Use the Autopilot Charts functionality to generate a visual representation of your data in the form of bar charts, line charts, and pie charts. Note that Autopilot Charts supports defects.

Query examples:

• Show me attachment distribution by file types.
• How many requirements are fully, partially, or not tested?
• Show failed test cases grouped by requirement.
• Show failed tests without linked defects grouped by requirements.
• Show test case distribution by label.
• Show distribution of requirements with custom field 'Sprint'.
• Which test sets take the longest to execute?
• Show me trends of testcases that haven’t been executed recently.
• Generate a pie chart with all open issues grouped by severity.
• Generate a bar chart with one bar for each requirement that is affected by Open issues, where each bar is a stack of defects grouped by their severity, having top severity on top.

Figure 10. Autopilot Charts in Autopilot Chat

AI-powered Requirement Evaluation

Use the Evaluate Quality functionality to assess the clarity, completeness, and testability of requirements, ensuring higher-quality inputs before test design begins.

To invoke a requirement evaluation from Autopilot Chat, reference the requirement by name or ID. Autopilot Chat understands the context, opens the Requirement Evaluation interface, and displays a detailed analysis with improvement suggestions.

After the requirement evaluation is completed, review the outcome, check the identified issues, and refine the requirement without leaving the chat.

Query examples:

• Evaluate the quality of the ‘Submit Loan’ requirement.
• Evaluate the quality of UIB:24.

How to use Autopilot Chat for requirement evaluation (Edit mode)

1. Enter a query indicating which requirement needs to be evaluated.

Figure 11. Autopilot Chat - Evaluate Requirement query

2. Select Configure and edit the fields: add any documents or select which documents to be included in the analysis, add or edit a prompt, select an AI model, and then either Accept or Reject the operation.

Figure 12. Autopilot Chat - Evaluate Requirement configuration

3. If you select Reject, the operation is not performed. If you select Accept, Autopilot works behind the scenes to provide the results.

Figure 13. Autopilot Chat - Evaluate Requirements result review

4. (Optional) Enter a query about the most immediate action that needs to be performed next.

AI-powered Test Generation

Use the Generate Test Cases functionality to automatically create high-quality, structured test cases based on requirement details, user documents, RAG or user prompts.

To invoke Generate Test Cases from Autopilot Chat, reference a requirement by name or ID. Autopilot Chat interprets your intent, identifies the relevant requirement, and launches the Generate Test Cases tool where you can provide more context (labels, custom fields).

After the tests are generated, review the generated tests.

Query examples:

• Generate test cases for the ‘Submit Loan’ requirement.
• Generate tests cases for UIB:24.

Leverage RAG vs. Attach documents

Leverage RAG only when:

• Requirements contain ambiguous terms or abbreviations
• Need to pull the ambiguous context from multiple sources dynamically

This is an example when you should leverage RAG: “Enable SSO for the CR, BH and FI modules and sync with the BI dashboard.” In this example scenario, documents should be uploaded for RAG, referencing the ambiguous abbreviations.

Attach documents when:

• Requirements do not contain ambiguous terms
• The files contain crisp information about the requirement
• You can define custom prompts and instruct LLMs to extract the context which is relevant for the requirement

How to use Autopilot Chat for test generation (Edit mode)

1. Enter a query which generates tests for a specific requirement.

Figure 14. Autopilot Chat - Test Generation query

2. Enter a query indicating which requirement needs to be evaluated. 3. Select Configure and edit the fields: add any documents or select which documents to be included in the analysis, add or edit a prompt, select an AI model, and then either Accept or Reject the operation. 4. If you select Reject, the operation is not performed. If you select Accept, review the results.

Figure 15. Autopilot Chat - Generate Test result review

Documentation Search

Ask questions in natural-language questions to learn how to use Test Manager. Autopilot Chat retrieves the information directly from the official documentation, complete with source links for further reading.

This makes onboarding new teams much faster and eliminates the need to switch between documentation and product screens.

Query examples:

• How do I create a requirement?
• How do I execute a test set?

Figure 16. Autopilot Chat - Documentation Search

Obsolete Test Detection

Use the Find Obsolete Tests functionality to maintain a clean, up-to-date test repository by automatically identifying the outdated or redundant test cases linked to requirements.

To invoke Find Obsolete Tests from Autopilot Chat, reference a requirement by either a name or ID. Autopilot Chat interprets your intent, identifies the relevant requirement, and launches the Find Obsolete Tests tool where you can provide more context.

Autopilot Chat analyzes the relationship between requirements and their associated test cases to detect obsolete test cases caused by:

• Updated or deprecated requirements
• Redundant coverage of the same functionality
• Outdated test environments or dependencies
• Misaligned or unsupported test steps

Finding obsolete test cases allows testers to focus only on relevant, executable test assets, improving both test accuracy and maintenance efficiency.

Figure 17. Autopilot Chat - Find Obsolete Tests

Web Reader and Web Search

When a query falls outside the scope of documentation or Autopilot Search, Autopilot Chat automatically switches to web search mode, but only within the context of Test Automation.

Autopilot Chat searches trusted public sources for topics related to testing frameworks, automation strategies, or QA methodologies, and then summarizes the most relevant insights.

Users receive contextual, automation-focused guidance while staying within the Test Manager domain.

Query examples:

• What are the best practices for writing automated regression tests?
• How should I design data-driven test cases in automation?

MCP servers and settings

MCP stands for Model Context Protocol — a modern standard that allows AI systems (like Autopilot Chat) to securely connect to external data sources, tools, and automation workflows.

MCP works as a bridge between your AI assistant and the systems or APIs it needs to interact with. Instead of the AI just generating text responses, MCP lets it take action — such as creating requirements, assigning test cases, or running a UiPath process — all through well-defined, secure interfaces.

An MCP server hosts a set of tools (or “capabilities”) that the AI agent can call when it detects relevant intent in your chat prompt.

Each MCP tool behaves like an API endpoint but is AI-aware — meaning the chat understands when and how to invoke it, based on natural language.

• You type a natural-language request in the chat, such as: “Import the last two requirements from this Excel file.”
• The chat interprets your intent and looks for an MCP tool that can perform this action.
• The relevant MCP tool (e.g., Create_requirement) is triggered, taking the input from your context (like the Excel data).
• The action is executed automatically — for instance, creating requirements or linking test cases.

In addition to the built-in Test Manager MCP server, only the MCP servers created in Orchestrator are supported. To learn more about the MCP servers which can be configured via Orchestrator, refer to Orchestrator - About MCP Servers and Orchestrator - Managing MCP Servers.

Open the Autopilot Chat settings and configure the MCP servers as described in the Autopilot - User interface topic.

MCP server types

• The built-in Test Manager MCP server provides the following tools:
  • Assign test cases to requirement
  • Assign test cases to test set
  • Assign object label
  • Bulk add labels
  • Bulk create test cases
  • Bulk set custom field values
  • Create requirement
  • Create test case
  • Create test set
  • Get all labels
  • Get all object labels by object ID
  • Get assigned test case IDs for requirement
  • Get custom field label values
  • Get custom field text values
  • Get requirements
  • Get test cases
  • Get test sets
  • Get test steps for test case
  • Get test executions
  • Get test case logs by test execution
  • Search objects by custom field value
  • Unassign test cases from requirement
  • Unassign test cases from test set
• The custom MCP servers created via Orchestrator expose your own processes, automations, or APIs — such as a workflow that generates test data or creates accounts in a third-party app.

Link to the Test Manager MCP server

You can configure the Test Manager MCP server in the Autopilot Chat settings.

1. Open the Autopilot Chat settings.
2. Under MCP Servers, enter a Test Manager server URL which observes this pattern: https://{yourcloudaccount.com}/{ACCOUNT_NAME}/{TENANT_NAME}/agenthub_/mcpfps/testmanager. Add your cloud URL, account and tenant name in the relevant placeholders. Example: https://cloud.uipath.com/TestCloudAccount/DefaultTenant/agenthub_/mcpfps/testmanager
3. Select Add and Save.