In 2025, Model Context Protocol (MCP) became arguably the most hyped technology in the AI and API community as a means to effectively connect large language models (LLMs) with tools, data, and APIs. The excitement is astounding, and MCP continues to dominate industry discussion and new feature releases. Yet, proven success stories are rare. So, does MCP deserve the praise?
In this keynote, I’ll present a well-rounded, unbiased take on MCP, covering both the good and the bad. We’ll consider when using MCP actually makes sense (and when it doesn’t), exploring strong case studies of its use in practice, as well as the potential downsides (such as over-engineering, security risks, and maintenance hurdles).
Attendees will take away a basic 101-level understanding of how MCP works and how it can be utilized to connect AI agents with underlying APIs. In broad strokes, I’ll explore the ecosystem that has risen around MCP, emerging best practices, and how it fits into the pre-existing API tooling landscape. Lastly, I’ll highlight some alternative formats and proposed standards for agentic connectivity.
Whether you’re MCP-curious or already building an MCP server, this opening talk will hopefully set the tone for how we think about agent-to-API interactions over the course of the two-day conference, and beyond.