Create a Task Agent

Navigate to the Agentic Workspace

In your project sidebar, click Agentic Workspace to expand it, then click Agents. You land on the Agents tab — this shows all agents in the project.

The page heading reads Agentic Workspace with three tabs: Agents, MAS Systems, and Agent Personas. Below those tabs is the Agent Management section with a search box, Filter/Sort buttons, and the Create Agent button.

Click Create Agent

Click the Create Agent button in the top-right of the Agent Management section. A dialog opens asking you to choose the agent type.

Choose the agent type

The dialog shows two cards:

Each card lists key features:

• AI Agent — Natural language configuration, A2A Protocol compatible, Dynamic tool selection
• Task Agent — Visual graph editor, Step-by-step control, Branching & loops

Click the Task Agent card. The wizard advances to the first step.

Step 1 — Basics (required)

The wizard header shows three steps: Basics → Input/Output (Optional) → Behavior (Optional). You’re on Basics.

A “Task Agent” badge appears at the top, followed by the heading “Name your agent & describe its purpose” and the note “You can configure everything else in the Agent Editor after creation.”

Fill in:

• Agent Name (required) — Give your agent a clear name (e.g., “Support Ticket Router”). Minimum 2 characters.
• Description (required) — Describe what the agent does. Minimum 10 characters. Use the AI Generate button to auto-generate a description from the name.

Click Next to proceed, or continue to create with just the basics — the remaining steps are optional.

Step 2 — Input/Output (optional)

This step defines what data your agent receives and produces. The heading reads “Define Input & Output” with the subtitle “What data does your agent receive and produce?”

Choose from six presets:

Click a preset to select it, then click Next. Or click Skip to leave schemas undefined (you can add them later in the Agent Editor).

Step 3 — Behavior (optional)

The final step configures the runtime. The heading reads “Runtime Configuration” with the subtitle “Choose which LLM and policy to use when this agent runs.”

Two sections:

LLM Configuration — Defaults to Project Default (“Uses whatever is configured at project level”). Click Change model to pick a specific model from the available options.

Policy (Guardrails) — Attach a policy to control agent behavior boundaries. If no policies exist in your project, a message directs you to create them in Configuration → Policies.

A Ready to Create summary card at the bottom shows your selections:

• LLM Config → Project Default (or your chosen model)
• Policy → None (or your chosen policy)

Click Create Agent to finish.

You’re on the canvas

After creation, qRaptor redirects you to the Agent Builder — a full-screen visual editor at its own URL. The agent is created as v1 • DRAFT.

Your canvas starts with two nodes already connected:

• Start (Manual trigger) — the entry point of your workflow.
• End (Exit) — the exit point.

The left panel shows the Node Palette with all available node types. The top toolbar gives you access to Variables, Validation, Undo/Redo, Testing, and Save/Publish controls.