Creating custom agents and using them in workflows

Beyond the catalog, CloudRadial AutomationAI lets you build your own agents and use them in workflows. This article covers creating a custom agent, generating one with AI, importing and exporting agents, defining an agent's input and output contract, and referencing an agent from the Agent node in the Designer. It is for workspace users who build workflows (Admin or Owner Role).

• Custom agents
• Creating a custom agent
• Generating an agent with AI
• Importing and exporting agents
• Input and output contracts
• Using an agent in a workflow

Custom Agents

The Custom tab on the Agents page holds the agents you author yourself. A custom agent can be created inline within a workflow for one-off use, or saved so you can reuse it across workflows by name.

Creating a Custom Agent

From the Custom tab, create a new agent. You define its goal, the extensions whose tools it may use, and its configuration. Save it to reuse it, or keep it inline to a single workflow.

Generating an Agent with AI

You can describe the agent you want in plain language and have AutomationAI generate it. The generation is grounded on your workspace's real extensions and workflows, so the draft agent references things you actually have. Review and adjust the generated agent before saving it.

Importing and Exporting Agents

Agents can be exported as YAML and imported from YAML, so you can share an agent, back one up, or move it between workspaces. Download an agent from the Custom tab to export it, and import a YAML file to add one.

Input and Output Contracts

An agent can declare an input schema and an output schema (as JSON schemas). The input schema defines the structured fields the agent expects; the output schema defines the structured result it returns. These contracts are what let a workflow pass data into an agent and read specific fields back out reliably.

Using an Agent in a Workflow

Add an Agent node to a workflow in the Designer and point it at an installed or custom agent by its slug. When the agent declares an input schema, the node shows a binding table where you map each input field to an upstream node's output. The agent's declared output fields appear as bindable paths, so downstream nodes can reference the agent's result. See the article on the Agent node for the full node reference.