Skip to main content

Agent Builder

Agent Builder is the intelligence layer of ART, designed to empower developers to create, configure, and deploy AI-driven agents through a simple, prompt-based setup, eliminating the need for complex coding or infrastructure management.

Agents built through Agent Builder work closely with MCP Servers and the Orchestrator to extend application logic beyond static flows, enabling dynamic, intelligent, and adaptive behavior in message pipelines.

Key features​

Agent Builder brings together flexibility, intelligence, and ease of use:

  • Prompt-Based Creation: Build agents in just a few steps, and no advanced coding required. Developers and non-technical users alike can design agents directly.

  • Seamless Integration: Connect agents to your existing systems securely. Extend capabilities with custom MCP servers for domain-specific knowledge.

  • Orchestrated Workflows: Enable multiple agents to collaborate, share context, and adapt dynamically. Transform complex business processes into streamlined automation.

How Agent Builder Works​

Agent Builder abstracts away the technical complexities of integrating with LLMs, APIs, and internal systems, allowing developers to focus purely on behavioral logic and intent definition through prompts.

1. Agent Creation​

Agents are defined using prompt-based configurations. Each agent can be trained or instructed to perform specific functions such as analyzing data, generating responses, making predictions, or calling tools via connected MCP Servers.

2. Integration with MCP Server​

Once created, agents connect with MCP Servers to gain access to real-world tools and application logic. The MCP Server acts as the execution bridge for the agent, translating its plain-language commands into actionable API calls.

When an agent decides to “perform” a function (for example, “fetch client transaction details”), the MCP Server handles:

  • Tool discovery (from the Manifest)

  • Authentication and secure execution

  • Result formatting and return to the agent

This partnership between Agent Builder and MCP Server allows developers to build intelligent systems that can reason, decide, and act, all from a natural-language interface.

3. Deployment in Orchestrator Flows​

After agents are built, they can be integrated directly into the Orchestrator as part of a channel’s message flow. When orchestration is enabled on a channel, messages passing through can be routed to an agent.