Studio Web - Agents

Autopilot agent in Studio Web for building, configuring, and evaluating intelligent agents.

Autopilot helps you build, configure, and evaluate intelligent agents in Studio Web using natural language.

Verwenden Sie den Chat für Folgendes:

• Design agent behavior by creating and updating system prompts, user prompts, and input/output schemas.
• Add and manage tools including solution processes, built-in capabilities, and third-party integrations.
• Attach knowledge contexts and indexes to give agents domain-specific expertise.
• Create evaluation sets and generate test cases to validate agent behavior and improve reliability.
• Retrieve and analyze the complete agent configuration to identify gaps and optimize performance.
• Refine tool and context descriptions so the agent makes better decisions about when to use them.

Verbundene Quellen

Agent configuration tools

• Get Agent Configuration—Retrieves the complete current agent configuration including system prompt, user prompt, input/output schemas, tools, contexts, and escalations.
• Update System Prompt—Replaces the agent's system prompt with new global instructions that define role, capabilities, and behavior patterns.
• Update User Prompt—Replaces the agent's user prompt with new per-request instructions that define how inputs are processed and outputs are structured.
• Update Input Schema—Replaces the agent's input schema with a new JSON Schema definition for structured inputs.
• Update Output Schema—Replaces the agent's output schema with a new JSON Schema definition for structured outputs.
• Update Tool Name—Renames an existing agent tool without changing its underlying functionality or input schema.
• Update Tool Description—Updates the description of an existing agent tool to improve guidance on when and how to use it.

Tool management

• Add Agent Tool—Adds a solution resource of type process, agent, or API as a callable tool available to the agent at runtime.
• Add Agent Built-In Tool—Adds a pre-configured internal platform tool such as file analysis, attachment loading, or batch transformation to the agent.
• Add Agent Context—Adds an index resource as a knowledge context that the agent can query for domain-specific background information.
• Search Integration Service Activities—Searches for available third-party Integration Service activities from connectors.
• Add Integration Service Activity—Adds an Integration Service activity as a placeholder tool that enables integration with external services.

Evaluation tools

• Create Eval Set—Creates a new evaluation set with custom name, batch size, timeout, and memory settings for organizing tests.
• Generate Eval—Generates and adds a new evaluation to a set based on the current agent definition with input generation instructions and expected behavior.
• Get Existing Evaluations—Retrieves existing evaluations from a set to understand covered scenarios and ensure mutual exclusivity.
• List Eval Sets—Lists all available evaluation sets in the current project with details and evaluation counts.

Knowledge and search tools

• Web Reader—Extracts readable text from a publicly accessible URL and optionally summarizes the content.
• Web Search—Performs a web search and returns structured results including titles, URLs, and snippets.
• UiPath Documentation Search—Searches UiPath official documentation and returns structured results with titles, URLs, and snippets.

Project tools

• Create Projects—Creates new empty projects in the current solution with specified names and project types.

For details, refer to About Autopilot for Agents and Autopilot for Agents tools.