From MCP to Multi-Agents: The Evolution of Agentic AI (and What's Next)
Mike joins the AWS Developers Podcast to discuss the evolution of agentic AI, MCP challenges, lightweight frameworks, and what's next for AI agents.
I joined the AWS Developers Podcast for a conversation about where agentic AI is heading. We covered a lot of ground in just over an hour — from the evolution of the Model Context Protocol (MCP) and its real-world challenges, to why lightweight agent frameworks like Strands are gaining traction, to the security concerns that come with giving agents access to your filesystem and command line.
We also talked about AI coding assistants as arguably the biggest success story in the agent space so far, context overloading as a practical bottleneck, and why 2025 is shaping up to be “the year of agents.”
Listen to the full episode here.
Topics Covered
- The Model Context Protocol (MCP) — where it’s working and where it’s struggling
- AI coding assistants as the primary agentic success story
- Security concerns with agent filesystem and command-line access
- The shift from prompt engineering to model-centric agent frameworks
- Context overloading challenges in production agents
- Lightweight frameworks like Strands Agents
- Multi-agent architectures and what’s coming next