Agents

Note: Availability of this feature depends on your license. For information about your license, reach out to your Nintex Account Manager, and for information trialing this feature, see Free Trials.

Use the Agents capability in Nintex CE Platform to create and manage agentflows. Agentflows can use AI agents to interpret information, reason, make context-based decisions, and generate outputs to complete specific tasks. Design agentflows in Agent designer to configure and manage its actions and logic. Each agentflow can incorporate one or more Agent actions, together with logic, operations, connector actions, and other actions that determine how the agentflow runs. For more information, see Agent designer.

To learn more and effectively use Agents, enroll in this Nintex University course (login required):

Jump to:

Understand key Agents terms

The following terms describe different parts of the Agents capability.

  • Agents: The umbrella term for the Nintex CE Platform capability used to create, run, manage, and evaluate agentflows and agent actions.

  • Agent designer: The drag-and-drop canvas used to create and configure agentflows. For more information, see Agent designer.

  • Agentflow: A The workflow containing one or more AI Agent actions which process you create, configure, save, publish, run, manage, and evaluate in the Agent designer. An agentflow can contain one or more agents and other actions. For more information, see Create an agentflow.

  • Agent action: The action you add to an agentflow to configure and run an agent. The action defines the model, connection, role, instructions, tools, knowledge, and output for the agent. For more information, see Agent.

  • Agent: An AI component configured and run through an Agent action in an agentflow. An agent interprets inputs, follows instructions, uses its configured tools and knowledge, and generates an output.

Creating an agentflow is not the same as configuring an agent. You first create an agentflow as the complete process and then add one or more Agent actions to configure the agents that perform AI-driven work within that process.

Access the Agentflows page

The Agentflows page displays the agentflow assets available to you.

  • Go to Agents > All agentflows

The Agentflows page is displayed. For more information, see Agentflows list.

What you can do in Agents

Use the Agents section of Nintex CE Platform to:

  • Create, open, and manage agentflows.

  • Build agentflows in the Agent designer.

  • Add and configure agents within an agentflow.

  • Test, save, and publish agentflows.

  • Review agentflow instances and their execution details.

  • Evaluate agentflow results against expected outcomes.

  • Manage supporting resources used by agentflows, such as datasets and connections.

The options available depend on your permissions and the features enabled in your tenant.

Create an Agentflow and its Agents

An agentflow coordinates Agents and Agent actions required to complete a task. When you work with Agents and Agent actions in an agentflow, follow this order of operations:

  1. Create an agentflow in Agent designer.

  2. Configure how the agentflow starts.

  3. Add one or more Agent actions.

  4. Configure the model, role, instructions, tools, knowledge, and output for each agent.

  5. Add logic, operations, connector actions, or other supported actions to control the process.

  6. Test, save, and publish the agentflow.

Each Agent action runs a single agent. An agentflow can use multiple Agent actions when different agents are required to perform separate tasks or collaborate as part of the same process. For more information about configuring an agent, see Agent.

Workflows vs. Agentflows

Workflows and Agentflows both automate processes, but they are designed for different types of work.

Consideration Workflow Agentflow
Primary purpose Automate repeatable business processes. Orchestrate flexible, AI-driven tasks.
Behaviour Follows predictable, predefined steps and rules. Uses agents to interpret information, reason, and make context-based decisions.
Logic Explicitly defined by the workflow designer. Combines explicitly configured actions with reasoning inferred by agents.
Use of AI Optional or assistive. Core to the process.
Control Provides greater predictability and control over each processing step. Provides greater flexibility and adaptability when inputs or required outputs can vary.

Use a workflow for tasks that follow known steps and rules and should produce predictable results. For example, use a workflow to:

  • Route an approval.

  • Send notifications.

  • Update records.

  • Assign tasks.

  • Apply fixed business rules.

Use an agentflow for tasks that require AI to interpret information, reason, generate content, or make context-dependent decisions. For example, use an agentflow to:

  • Interpret an unstructured customer request.

  • Classify content based on its meaning.

  • Extract or summarize information.

  • Generate a response or report.

  • Assess information and recommend an action.

You can also use workflows and agentflows together. A workflow can manage the wider business process and call an agentflow when a specific step requires AI-driven processing. For example, a workflow can receive and route a customer request. It can Call an agentflow to classify the request and draft a response before continuing with an approval or notification step. Agentflows are maintained as separate assets so that AI-driven processes can be reused by multiple designs. This also allows an agentflow and its agents to be maintained, governed, tested, and evaluated independently.

For an introduction to workflows, see Workflows and the Quick start guide.