Design Assistant
Introduction
The Design Assistant is an AI-based chatbot integrated into the Solution Designer. It supports users during their daily design work by answering questions about the current project and executing design tasks on their behalf.
The assistant is accessible from anywhere within the workbench, so teams can get guidance quickly without breaking their flow.
Prerequisites
To use the Design Assistant, generative AI credentials must be available. Your administrator can configure a central AI provider connection via the k5-ai-credentials secret — see AI providers. Alternatively, you can provide your own API key for OpenAI platform or OpenAI on Azure in User settings.
Supported LLM providers when configured centrally:
- OpenAI platform
- OpenAI on Azure
- Amazon Bedrock
Capabilities
The Design Assistant currently supports:
- Answering questions about the current design — explore and understand the model elements, relationships, and overall structure of your project through natural-language questions. This includes what is drawn on a diagram you have open or name, and external proxy elements, which represent objects whose original data is managed in an external system.
- Executing design tasks — create, edit, and delete model elements and relationships, and perform other design operations based on your instructions. See Create, edit, or delete relationships for what the assistant can change on a relationship.
- Diagram content — ask what a diagram shows, and place elements as nodes, draw relationships as edges, update how they are drawn, or remove them from the diagram. Taking a node or edge off a diagram leaves the element or relationship in the project.
- Clarification questions — when a request can be carried out in more than one way, the assistant pauses and shows the choices as option boxes in the chat. Click a box to answer, or type your own answer when the card includes a text field.
- MCP support — call tools from MCP servers that administrators registered as external systems. Before a tool runs, the assistant asks you to allow that call. Read and writing tools are available unless an administrator restricts the assistant to read-only tools. Combined with external proxy elements, the assistant can fetch information from a connected system and keep a representation of it in the model. For a complete walkthrough with Jira (Rovo MCP and Basic authentication), see the Jira integration example.
In the chat you can also @ mention a project, diagram, model element, or MCP external system so the assistant focuses on that reference. Mentions are design and system references, not @-mentions of people.
With cross-project support, the assistant can read from another project or branch you refer to (via mention or by name) and use that information while designing in the current project. Write operations remain limited to the current project and branch.
You can keep multiple chat sessions open in parallel as tabs, each with its own history — useful when you work on several design topics at once. Within a session, the assistant remembers what you asked for and what it already changed.
Design recipes extend the assistant with skills and commands. Recipes linked to the project baseline apply for the whole team; you can also upload personal design recipes in User settings so your enabled skills and commands apply on every query, independent of the baseline.
The capabilities of the Design Assistant are limited to certain tasks in this release. Support for additional operations will be added in upcoming versions.
Content generated and persisted by the Design Assistant is marked as AI-generated in a machine-readable way.