Talk
Building with AI Agent Protocols
The agentic AI landscape is rapidly evolving from isolated experiments to production-ready products, but a critical challenge remains: how do we build solutions that enable AI agents to reliably interact with tools, services, and each other? We've heard about MCP, ACP, A2A… but which protocols bring leverage to building solutions today? This talk explores the emerging standards that product teams are using to build real user-facing solutions. Model Context Protocol (MCP): The open standard for connecting AI systems to external data and tools—with practical lessons from shipping MCP-powered features. MCP Apps: How integrating user interfaces directly into the agent experience is turning agents from backend utilities into products users actually interact with. Agent Communication Protocol (ACP): Standardizing how a frontend can interface with a stateful, asynchronous agent backend. Drawing from real-world product development, we'll cover architectural patterns that work in production, performance considerations that affect user experience, and the path from prototype to shipped product.
About
Brad is Tech Lead for AI at Block. He built and maintains goose, the open source AI agent framework owned by the AAIF that automates repetitive tasks for thousands of engineers in the company. He also leads company-wide initiatives to optimize data and ML workflows and evolve platform architecture. Brad has a long history in Python, distributed compute systems, and applied machine learning. Brad previously earned his physics PhD from UC Berkeley and worked on processing and analyzing data at the Large Hadron Collider at CERN in Geneva.