Agent Builder

4.1. Overview

Agent Builder is a visual platform for designing, testing, deploying, and operating AI-powered workflows. It combines a browser-based workspace, reusable components, runtime APIs, flow execution services, credential management, and integrations with model providers, vector databases, storage systems, external business services, and Python RPA robots.

The platform supports more than 150 components across models and agents, input/output, flow control, processing, tools, data sources, knowledge bases, embeddings, LLM operations, cloud integrations, vector stores, document processing, RPA, and utilities.

<aside> 💡

Agent Builder flows can be used interactively in the visual editor or exposed through the Runtime API for applications, schedulers, RPA processes, and external systems.

</aside>

4.2. Key Platform Capabilities

4.3. Building a Flow

  1. Add input components such as Chat Input, Text Input, or Webhook.
  2. Add a model or agent and configure its provider credentials.
  3. Connect prompts, memory, knowledge sources, processing components, and tools.
  4. Use flow-control components such as If-Else, Loop, Run Flow, or Data Conditional Router where required.
  5. Add Chat Output or Text Output and test the complete flow.
  6. Review credentials, tool permissions, error handling, and logs before deployment.

4.4. Components and Tool Mode

Components are reusable workflow building blocks with typed inputs and outputs. Tool-capable components can be exposed to an Agent as callable tools. In Tool Mode, a component returns one output of type Tool.