Using the Agent

The pConAI agent is designed to understand natural language intents and translate them into precise architectural actions. This guide explains how to interact with the agent effectively.

The Chat Interface

When you open the pConAI web interface or your integrated client, you'll be greeted by the chat prompt. Here, you can type your requests in plain English (or your preferred supported language).

Best Practices for Prompting

To get the best results, follow these guidelines:

1. Be Specific: Instead of saying "add a chair", say "add the Aeron office chair by Herman Miller in black mesh".
2. Define Boundaries: When creating spaces, mention dimensions. "Create a 5x5 meter conference room with a glass wall on the south side."
3. Chain Commands: You can ask the agent to perform multiple steps at once. "Insert a desk, place a laptop on it, and adjust the camera to an isometric view."

Capability Scopes

Depending on your subscription tier (Pilot, Studio, or Enterprise), the agent may ask for confirmation before executing large-scale structural changes to your scene.

If a command fails or requires disambiguation (e.g., if there are 3 different "Aeron" chairs in the catalog), the agent will present you with a list of options to choose from. Simply click the desired option or type its name to proceed.

[!WARNING] Destructive actions like Delete all items or Clear scene will always require explicit confirmation to prevent accidental data loss.