App Builder AI Agent

We have started developing an AI agent for App Builder and are making it available to early testers. The AI agent should still be considered beta, and it may still change considerably. We are excited to receive your feedback via our forum, community catchups, or onboarding calls.

First example

Check out the Urban Frame App, an example where the agent is used to drive the user experience around a small district planning application.

Agent setup

You can find a link to the AI Agent settings in the Developers settings of the platform, or directly at the following link: https://agent.shapediver.com/settings (you need to authenticate to your ShapeDiver account to open the link).

Create a profile

You can create one or several profiles for the agents you include in your ShapeDiver Apps. A profile is a set of settings for the Agent, most importantly the settings needed to link the LLM provider of your choice.

Set a default profile

After creating a first profile, you can set it as default: it becomes the profile used by your Apps without further setup.

Choosing a profile

If you want your App to make use of a different profile, you need to define a profileName property for your agent in Grasshopper. The name you specify refers to the name of the profile you defined in the AI agent settings.

General profile settings

• Name: The name of this profile. Used to reference the profile in specific Apps.

• Description: An optional description for this profile.

• Extra system prompt: An optional system prompt to be included in the context of the Apps making use of this profile. Note that each App is able to define its own lower-level system prompt.

• Reasoning: Three visibility levels for the LLM’s thinking steps.

  • Show: Expand while reasoning streams, then collapse.

  • Collapse: Keep the Thinking block collapsed until the user opens it.

  • Hide: Do not show reasoning in the chat.

• New thread prompt: Prefilled in the message box when a new thread is created. Switching to an existing thread or forking leaves the box empty. Clear this field to start new threads blank.

LLM Provider settings

• LLM Provider: Choose between a list of supported providers. The current list include the following providers, and it will be expanded in the future:

  • openai

  • openrouter

  • foundry

  • bedrock

  • anthropic

• API Key: The key used to link your chosen LLM Provider. The key is securely stored on the backend.

MCP Servers

You can configure MCP servers the agent should have

A2A Servers

TODO

Observability settings

TODO

Getting started

Once at least one profile is configured, you can start including agents in the Apps you create using the App Builder.

Add a generic Agent to existing Apps

On the one hand, it is possible to include an agent in existing Apps without any changes to the Grasshopper definition. TODO

Define a specialist Agent in your definitions

In order to use the full potential of the ShapeDiver Agent framework, we recommend defining agents directly inside your Grasshopper definitions, using the Agents input of the App component. This will allow you to control context efficiently and define custom tools for the agent to interact with the model.

When an agent is included in your definition, no further setup is required; the standard URL of the model will already included the Agent widget.

We will soon add more detailed documentation about the Grasshopper workflow, but you can already find a first tutorial model here to get started.

Agent tools

TODO

Limitations

Limited parameter types

The AI agent currently supports the following parameter types:

• Number (float, integer, even integer, odd integer)

• Boolean

• String (see Direct Text Input)

• String list (value lists)

• Color

Support for further parameter types like File will be added.