Agent Crew

What is Agent Crew in ZBrain Builder?

Agent Crew is a multi-agent orchestration capability within ZBrain Builder that enables enterprises to execute complex, multi-step tasks through the coordinated efforts of multiple AI agents. Instead of relying on single agents for isolated actions, Agent Crew allows organizations to design, manage, and monitor interconnected agents that work in a structured, hierarchical setup to achieve end-to-end task automation.

At its core, the Agent Crew framework introduces a new coordination model—where a supervisor agent governs one or more subordinate/child agents, defining the logic for how each task is delegated, executed, and validated. This modular hierarchical approach simplifies complex automation scenarios by breaking them into manageable, role-specific tasks, improving both transparency and control.

Core concepts

Supervisor agent

The supervisor agent acts as the central decision-maker within a crew. It:

• Receives the primary input for the entire workflow.

• Determines how the input should be processed.

• Delegates specific tasks to child agents based on pre-configured logic.

• Evaluates outputs from child agents and decides on the next action.

Child agents

Subordinate agents are task-specific agents configured to handle well-defined parts of the overall workflow. They:

• Operate autonomously within their scoped instructions.

• Use tools, functions, or integrations to perform parsing, API calls, or data extraction.

• Return results to the supervisor agent or perform follow-up tasks if part of a sequential chain.

Crew

A crew is a logical unit that includes the supervisor and its child agents. The crew structure:

• Defines how agents interact.

• Establishes dependencies between agents.

• Supports testing as a single coordinated unit.

Architecture highlights

• Multi-agent hierarchy: Agents are structured in a supervisor-child format, enabling the distribution of tasks and the logical separation of responsibilities.

• Prompt-governed delegation: Supervisors and child agents operate based on defined instructions (prompts) that guide task allocation, execution logic, and communication flow within the crew.

• Tool-based execution: Agents can use JavaScript or Python tools to perform specific tasks, such as parsing, transformation, API calls, or validations. These tools are reusable across agents and workflows, allowing better maintainability and consistency across automation logic.

• External system integration via MCP: MCP servers enable secure connectivity to enterprise systems like databases, APIs, CRMs, and email services. Once defined, MCP settings (authentication, headers, endpoints) can be used across multiple agents and crews.

• Failure handling logic: Retry rules and fallback instructions can be defined at the supervisor or tool level to ensure process continuity.

• Performance insights: Built-in dashboards report on token usage, success/failure rates, and execution durations to help teams identify bottlenecks and optimize agent design.

• Execution traceability: Logs for each agent’s action, inputs, outputs, and failures are captured independently, simplifying debugging and optimization.

Orchestration frameworks

ZBrain Builder agent crew supports multiple orchestration frameworks, giving users flexibility in how multi-agent workflows are modeled:

• LangGraph: Supports stateful, graph-based agent execution. Useful for workflows with complex branching logic and persistent context across agent interactions.

• Google ADK: A modular, model-agnostic framework for building and deploying AI agents with flexible orchestration patterns, multi-agent hierarchies, and built-in evaluation—optimized for developer-friendly agent workflows.

Users can choose the appropriate framework based on how dynamic or deterministic their automation needs are.

Why use Agent Crew

Agent Crew is designed to simplify the automation of complex, multi-step workflows across functions like customer support, sales, HR, operations, and legal. Whether it’s customer onboarding, document processing, or internal research workflows, Agent Crew offers structured automation with control, visibility, and flexibility.

Key benefits include:

• Clear task allocation and execution logic: Multi-agent hierarchy allows responsibilities to be distributed logically—supervisors control flow, child agents handle specialized tasks. This clarifies execution flow and makes it easier to locate and address task-specific failures.

• Reusable components: Tools and MCP configurations are modular and shareable across agents. This reduces duplication, shortens setup time for new agents, and improves long-term maintainability.

• Secure system integration: Through MCP, agents can safely interact with enterprise systems—databases, CRMs, APIs, file stores—without embedding credentials in prompts or tools. Centralized auth management ensures compliance and auditability.

• End-to-end observability: Built-in monitoring features—execution logs, I/O tracing, retry handling, and dashboards—provide transparency and insight into agent activity, token usage, errors, and success rates.

• Faster workflow updates: Logic changes can be made at the agent, tool, or prompt level independently. This modularity makes it easier to iterate, test, and deploy updates without breaking the full chain.

• Improved reliability and control: Supervisor prompts can include retry logic, timeouts, and fallback options. This ensures workflows remain operational even when certain agents fail or encounter edge cases.

• Adaptable to business-specific logic: Since each agent can be configured with custom tools, API logic, and MCP servers, crews can adapt to workflows with domain-specific validation, compliance checks, or decision criteria.

Summary

Agent Crew introduces structured, modular orchestration to enterprise-grade multi-agent systems. It enables a supervisor-subordinate setup where a central agent governs task flow and coordination, while specialized child agents execute individual steps. This architecture supports tool-based task execution, flexible orchestration logic, and integration with external systems through MCP.

With Agent Crew, enterprises can streamline complex tasks using intelligent agents that are observable, secure, and easy to manage at scale.