Agent Builder
Agent Builder is the Dataworkz no-code interface for creating, configuring, and testing Agents using plain-English descriptions and a guided UI.
For Business Users
Agent Builder lets you create a working AI Agent without writing any code or JSON. You describe what the Agent should do in plain English — for example, "Help customers track orders, initiate returns, and answer product questions" — and Agent Builder generates an initial configuration for you.
From there, you refine the Agent through a visual interface: add or edit Scenarios, connect data sources, set the Agent's persona, and test responses — all without leaving the browser.
For Developers
Agent Builder is a hybrid chat-and-form interface built on the same Dataworkz Agent framework that powers production Agents. It produces a standard Agent configuration (see Agents) that can be deployed via the Agent API.
Workflow
Describe the Agent — Enter a plain-English description of what the Agent should do. An underlying meta-Agent interprets this and generates an initial configuration, including suggested Scenarios and persona.
Review and refine — Edit the generated configuration in the UI. Adjust Scenario descriptions, add example queries, and update planning instructions.
Add Tools — Browse the Tools Repository and attach Tools to Scenarios. Each Tool represents a connection to a data source or capability.
Set context values — Define Agent-level variables (for example,
tenantId,defaultCurrency) that are available across all Scenarios.Test — Use the built-in chat panel to send sample queries. Inspect which Scenario was selected, which Tools were called, and what responses were generated.
Publish — Deploy the Agent to make it available via the Agent API or embed it in a product interface.
Key Interface Sections
Agent description pane
Chat interface for describing the Agent in plain English and iterating with the meta-Agent
Scenarios panel
List of all Scenarios; click any Scenario to edit its description, examples, Tools, and instructions
Tools panel
Browse and attach Tools from the Tools Repository
Context variables
Define shared variables available to all Scenarios
Test / Run panel
Live chat interface for testing the Agent before publishing
Real-World Example
A product manager at a SaaS company needs a support Agent for their help center. They open Agent Builder and type:
"Help customers troubleshoot login issues, reset passwords, and understand billing charges."
Agent Builder generates three Scenarios: LoginTroubleshooting, PasswordReset, and BillingInquiry. The product manager attaches a RAG Application Tool (linked to the help center knowledge base) to LoginTroubleshooting, a REST API Tool (linked to the identity service) to PasswordReset, and a Relational DB Tool (linked to the billing database) to BillingInquiry. After testing a few sample questions, they publish the Agent.
No code was written. The entire process took under an hour.
Related Concepts
Agents — The configuration structure Agent Builder produces
Scenarios — How Scenarios are structured and how routing works
Tools Repository — The catalog from which Tools are selected in Agent Builder
Tools — What a Tool is and how it is defined
Last updated