Client Service Orchestration Agent
This challenge tasks developers with building a sophisticated multi-agent system designed to orchestrate seamless customer support across diverse communication channels (e.g., chat, email, voice). The system will leverage CrewAI for role-based agent orchestration, with specialized agents communicating via an A2A (Agent-to-Agent) Protocol. Crucially, it will integrate with simulated enterprise systems (CRM, knowledge bases) using an MCP (server for secure, structured tool invocation. This project focuses on creating a 'supervisory' agent that routes inquiries to the most appropriate 'specialist' agents. These specialist agents (e.g., 'Technical Support Agent', 'Billing Agent', 'Sales Agent'), powered by Claude Opus 4.1 for nuanced understanding and response generation, will collaborate, share context, and utilize MCP-enabled tools to resolve complex customer issues efficiently. The challenge emphasizes secure, structured inter-agent communication and robust enterprise system integration.
What you are building
The core problem, expected build, and operating context for this challenge.
Orchestrate a multi-agent client service system using CrewAI, A2A Protocol, Claude Opus 4.1, and MCP for enterprise tool integration.
How work is evaluated
Evaluation focuses on the system's ability to accurately classify customer intent, route inquiries, perform multi-agent collaboration, correctly use MCP tools, and provide coherent, context-aware responses.
Challenge datasets
Public datasets, plus the private datasets you have access to.
What you should walk away with
Integrate Claude Opus 4.1 as the core reasoning engine for specialist agents, enabling complex natural language understanding and empathetic responses.
Design and implement a multi-agent system using CrewAI with clearly defined roles and tasks for customer service.
Develop and deploy an MCP (Multi-Agent Communication Protocol) server for managing and securing tool integration with simulated enterprise systems.
Build MCP-enabled tools that allow agents to interact with a simulated CRM, knowledge base, and ticketing system.
Orchestrate a complex customer journey involving multiple agent handoffs and collaborative problem-solving.
Implement A2A (Agent-to-Agent) Protocol for secure, structured communication and context sharing between agents.
Reference links and supporting material
Sample data for 2 tasks
Challenge setup
Evaluation focuses on the system's ability to accurately classify customer intent, route inquiries, perform multi-agent collaboration, correctly use MCP tools, and provide coherent, context-aware responses.
Challenge input
{'channel': 'string', 'customer_id': 'string', 'initial_query': 'string'}
Agent execution
No agent tool is listed for this challenge.
Evaluated output
{'final_response': 'string', 'agent_interactions': 'list of dicts', 'mcp_tool_calls': 'list of dicts', 'customer_satisfaction_score': 'int'}
- Checks if all necessary MCP tools for a given query were invoked.
- Verifies correct agent handoffs and context transfer via A2A protocol.
- ResponseAccuracy target: 90
- Python execution harness
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
No tool records are attached.
Evaluation contract
- Checks if all necessary MCP tools for a given query were invoked.
- Verifies correct agent handoffs and context transfer via A2A protocol.
Metrics
- ResponseAccuracy · Target 90
- ToolParameterCorrectness · Target 95
Workflow tasks
- Task 1
Input: {'channel': 'string', 'customer_id': 'string', 'initial_query': 'string'}
Output: {'final_response': 'string', 'agent_interactions': 'list of dicts', 'mcp_tool_calls': 'list of dicts', 'customer_satisfaction_score': 'int'}
Execution harness: Python
- Task 2
Input: {'agent_role': 'string', 'task_query': 'string'}
Output: {'mcp_tool_calls': 'list of dicts', 'response_from_tool': 'string'}
Execution harness: Python
Recipe freeze marker
No recipe freeze time is recorded.