- Getting started
- UiPath Agents in Studio Web
- UiPath Agents in Agent Builder
- UiPath Coded agents
Agents user guide
About coded agents
UiPath® coded are agents built through direct code development within a developer's preferred Integrated Development Environment (IDE). They provide developers with complete control over the agent logic, behavior, and integration with external systems.
.nupkg package, in a package feed of your choice. From there, it
can be used to create processes within one or more folders.
The coded agent is deployed as a UiPath process within Orchestrator folders. Once integrated in the platform, it adheres to the same governance principles as standard processes. This allows for organized deployment within folders, and means you can schedule, trigger, and monitor agent execution, while integrating it into broader automation workflows.
The following illustration outlines the process of creating, deploying, and executing a coded agent within the UiPath platform.
The UiPath SDK library and its supported languages and frameworks are continuously expanding. We currently support:
| Language | Repository | Functionality |
|---|---|---|
| Python | uipath-python |
Python SDK that features a CLI for creation, packaging, and deployment of agents to the UiPath Platform. It also enables programmatic interaction with UiPath Platform resources like processes, assets, buckets, Context Grounding, data services, jobs, and more. |
| Python with LangGraph | uipath-python extension: |
Python SDK that enables developers to build and deploy LangGraph agents to the UiPath Platform. This package provides programmatic interaction with UiPath PlatformTM services and human-in-the-loop through Action Center. |
For the latest information regarding UiPath SDK updates, refer to the Release policy.
Integrating UiPath resources via SDK
UiPath SDKs provide developers with the necessary logic to invoke and interact with other UiPath resources directly within their code, enabling a seamless integration with existing UiPath functionalities. For example, developers can use UiPath assets for secure storage and management of sensitive data, access and manipulate data stored in UiPath storage buckets, and establish and manage connections to various systems and services. Additionally, developers can trigger and interact with other UiPath processes from within the coded agent.
For details about the UiPath resources you can use in your coded agent, refer to the SDK documentation.
For more examples and implementation patterns, check out the sample projects in our GitHub repository.
Human in the loop for coded agents
You can define interrupt points within your agent to ask for human input. When an interrupt is triggered, the agent execution pauses, and the assigned user is asked for input. Upon human input, the execution is resumed.
How you define interrupt points varies depending on the agent language:
- Some frameworks, such as LangGraph, offer native interrupt patterns, allowing for human intervention in the agent's execution.
- With the UiPath SDKs, you can introduce a UiPath task as the interrupt
logic. When an interruption is triggered, the execution pauses, and a task
is created in Action Center.
- Users can review the agent's progress, provide feedback, or make decisions and act on the generated task via an action app.
- Once the human interaction is complete, the agent resumes its execution from the point of interruption.
Configuring environment variables for coded agents
You can configure your project with environment variables using secret assets, eliminating the need for direct code modifications.
Leveraging the Orchestrator support for integrating with external credential stores like CyberArk for credential assets, you can link to your CyberArk store and configure an environment variable for your agent to use that secret asset. At runtime, Orchestrator securely retrieves the key from CyberArk via the asset and makes it available to your agent as an environment variable.
Navigate to your agent's process in Orchestrator to define and manage API keys, runtime parameters, and other configuration values required by your agent's code. For details, refer to Starting an agent job.
Enabling tracing and debugging with LangSmith
LANGCHAIN_TRACING_V2, LANGCHAIN_API_KEY,
LANGCHAIN_PROJECT) to enable detailed tracing of your agent's
execution within LangSmith.
This integration provides comprehensive traces in your LangSmith account for every UiPath agent run. This allows you to analyze the agent's reasoning, execution path, and identify areas for debugging and performance optimization.
