Welcome to ACP: Agent Communication Protocol built in partnership with IBM Research's BeeAI. In this course you build and run agents that communicate through ACP, or the Agent Communication Protocol. You will configure sequential and hierarchical workflows of multi-agents that collaborate through a unified REST interface provided by ACP. I'm delighted that your instructors for this course are Sandi Besen, who is AI Research Engineer and Ecosystem Lead, as was Nicolas Renotte, who is head of AI Developer Advocacy at IBM. Thanks, Andrew. We're excited to work with you on this course. To build the team of agents that communicate and collaborate through ACP, you can host an agent on an ACP server, which will receive REST requests from an ACP client and forward those requests to the agent to execute. The ACP client, which can itself be an agent or any other process, discovers the agents using the endpoints of the ACP servers and then initiates request. This unified client-server interface helps standardize the communication between agents across teams. Take for example, a multi-agent system for customer support. One team might build a logistics agent to answer questions related to order status. A different team might build a different RAG agent to answer general questions related to products. Each team can make the agents available through an ACP server, while a the third agent hosted inside the ACP client can process customer inquiries and route them to the specialized agents. Each team can implement their agents using a totally different framework, or even switch between frameworks without needing other teams to make any changes in their own code. Because the communication happens through a standardized protocol. This standardization also makes it easy to integrate any existing agents into the multi-agent workflow. You can wrap an agent in an ACP server, which enables an ACP client to discover it and integrated into the workflow. These existing agents can also be made visible in a registry or agent catalog. The catalog can provide centralized agent listings and simplifies searching for agents, especially at large-scale deployments or enterprise Environments. ACP is an open source protocol that it's also been governed, meaning no single company controls it. Many open discussions with the community have shaped the development of this protocol and continue to shape its evolution. It's also worth noting that ACP and MCP can be seen as complementary protocols: and ACP agents can use MCP to access tools and then use ACP to interact with other agents. In this course you'll first wrap a RAG agent built using the CrewAI framework in an ACP server, and then interact with it through an ACP client. To create the sequential workflow you'll wrap a second agent built using the smolagents framework in another ACP server. Inside the ACP client, you'll learn how to discover the two agents and chain them sequentially. After that, you'll build another hierarchical workflow where instead of sequentially chaining the agents, you'll use a third routing agent coded using the smolagents framework in the ACP client, which will wrap the input queries to one of the specialized agents. You'll then extend the smolagents ACP server to use MCP to get access to tools. And finally, you'll learn how you can discover and run and import ACP agents to the BeeAI platform, which is an open source registry for agent discovery. Many people have worked to create this course. I'd like to thank from IBM, Kate Blair and Anna Fucs. From DeepLearning.AI Hawraa Salami also contributed to this course. ACP standardizes communication between sets of agents that might have been developed by the same developer or team, or developed by different teams in the same organization. But it's not only intended for local development environments, it can also be used to connect distributors to agents, even across different organizations. And that's specifically what you'll learn more about in the first lesson with Sandi. So let's go to the next video to get started.