The illustrations were created in cooperation between humans and artificial intelligence. They show a future in which technology is omnipresent, but people remain at the center.
AI-generated illustration
Developing AI agents using MCP, A2A, and ACP
Open Protocols for Tool Integration and Agent Interoperability: Architecture, Implementation, Security, and Best Practices
Online
2 days
German
€ 1.390,-
plus VAT.
€ 1.654,10
incl. VAT.
Booking number
42723
Venue
Online
.svg)
5 dates
€ 1.390,-
plus VAT.
€ 1.654,10
incl. VAT.
Booking number
42723
Venue
Online
5 dates
plus VAT.
incl. VAT.
Booking number
Venue
0 Dates
Become a certified
Machine Learning Engineer
Machine Learning Engineer
This course is part of the certified Master Class "Machine Learning Engineer". If you book the entire Master Class, you save over 15 percent compared to booking this individual module.
To the Master Class
In-house training
In-house training just for your employees - exclusive and effective.
Inquiries
In cooperation with
In cooperation with
In this hands-on training , you training how to develop modern AI agents that interoperate seamlessly with tools, systems, and other agents. The focus is on open protocols such as the Model Context Protocol (MCP) and agent communication standards like A2A and ACP.
You will develop a solid understanding of the architecture, design, and implementation of agent systems and apply this knowledge directly in practical exercises. In the process, you will design tool contracts, implement your own MCP server, and build a multi-agent scenario in which agents delegate tasks, exchange results, and coordinate processes.
A special focus is placed on security and governance. You will learn how to design agent systems that are controllable, traceable, and production-ready—from authentication and authorization to protection against prompt injection and misuse.
Contents
1. Fundamentals and Protocol Overview
- Challenges in tool integration and agent interoperability.
- Overview of MCP, A2A, and ACP.
- Application scenarios: tool integration, delegation, multi-agent systems.
2. MCP Design and Tool Contracts
- Role model: Host, Client, Server.
- Design of interfaces and data structures.
- Definition of inputs, outputs, and error cases.
- Development of robust and validatable contracts.
3. Implement the MCP server
- Structure and Architecture of an MCP Server.
- Tool registry and integration of functions.
- Validation, timeouts, and edge cases.
- Development of the first tools and services.
4. Integration into applications and workflows
- Tool discovery and function invocation.
- Processing structured results.
- Reliability through retries, rate limits, and logging.
- Establishing stable integration flows.
5. Security and Governance
- Trust boundaries and data classification.
- Authentication and Authorization (Least Privilege).
- Protection against prompt injection and misuse.
- Auditability and traceability.
6. Agent Communication with A2A
- Roles and division of responsibilities among agents.
- Communication patterns: Request, Status, Result.
- Task management and delegation.
- Handling long-running processes.
7. ACP and API Design for Agents
- Design of Capability APIs.
- Job lifecycle and state models.
- Error Handling Strategies and Idempotence.
- Mapping to agent communication.
8. Reference Architecture for Agent Systems
- Interaction between MCP, A2A, and ACP.
- Architecture of agent mesh systems.
- Versioning, compatibility, and testing.
- Governance, Policies, and Logging.
9. Developing Multi-Agent Systems (Capstone)
- Development of an end-to-end agent system.
- A combination of tools, agents, and workflows.
- Quality assurance, review, and documentation.
10. Production Readiness and Operations
- Requirements for production-ready systems.
- Monitoring, incident handling, and operations.
- Governance processes and rollout strategies.
Your benefit
- You'll learn how to develop AI agents in a structured way and integrate them into existing systems.
- You understand open protocols such as MCP, A2A, and ACP and apply them in practical situations.
- You develop your own tool integrations and build multi-agent systems.
- You can design robust, secure, and traceable agent architectures.
- You take security and governance requirements into account from the very beginning.
- You'll learn practical best practices for production-ready AI agent systems.
trainer
No items found.
Methods
The training concise theoretical sessions with an intensive practical component:
- Live demos and structured architecture overviews
- Guided hands-on exercises and code-along sessions
- Development of proprietary components such as MCP servers and APIs
- Final Project on the Implementation of a Multi-Agent System
An MCP-compatible training environment training provided for the training .
Final examination
Recommended for
This training for anyone who wants to develop AI agents and implement them in a business setting:
- Software and platform teams that are opening up applications to AI agents
- AI, data, and ML teams looking to integrate agents with data and tools
- Solution Architects and Tech Leads who design and evaluate agent architectures
- developers a background in APIs (e.g., Python, JavaScript, or TypeScript)
.svg)
