Agents

Drop pre-built agent nodes into any workflow — Standard, Reasoning, Autonomous, Multimodal, Domain-Specific, and more.

Overview

The Agents screen is the catalogue of all ready-to-use AI agent types in TotalApp. Each agent type encapsulates a specific set of behaviours, model preferences, and default system prompts. You drag agents from this catalogue into the Workflow Editor canvas without needing to configure the underlying model or prompt from scratch.

Agent Types

Standard Agent

General-purpose LLM call. Receives input text, returns generated text. The baseline building block for most AI pipelines.

Reasoning Agent

Uses chain-of-thought prompting to work through multi-step problems before producing a final answer. Ideal for analytical tasks.

Autonomous Agent

Operates in a ReAct loop — reasons about the task, selects a tool, observes the result, and repeats until the goal is achieved or a step limit is reached.

Multimodal Agent

Accepts images alongside text. Use for describing images, extracting text from screenshots, or visual Q&A pipelines.

Domain-Specific Agent

Pre-configured with specialised system prompts for a vertical domain (legal, medical, financial). Reduces the prompt engineering needed for narrow-domain tasks.

Customise Any Agent

Every pre-built agent type can be customised further via the Agent Architect. Open an agent from this catalogue in the Architect to change its model, adjust its system prompt, or attach additional tools from the Tool Registry.

Using Agents in Workflows

Open the Agents catalogue from the AI Orchestration sidebar.
Find the agent type you want and click Add to Workflow (or drag it directly to the Workflow Editor canvas if both screens are visible).
In the Workflow Editor, connect an upstream node's output handle to the agent node's input handle.
Configure the agent node in the right-side Config panel — set the task prompt, tool access, and knowledge base references.
Click Run to execute the pipeline.