EP 3 - MCP in Claude, AI Updates and Memory Deep dive into LangMem, Letta, MemOS and More.

In this episode of The Build, Cameron Rohn and Tom Spencer dig into the interplay of memory systems, agent architectures, and startup strategies that drive modern AI product development. They begin by unpacking MCP in Claude and a memory deep dive across LangMem, Letta, and MemOS, referencing Memos Paper while addressing authentication, API integration with Gmail, and containerized deployments via Docker. The conversation then shifts to developer tooling and infrastructure choices—Langsmith for orchestration, Vercel and Supabase for deployment and backend, and how Pulse MCP and other MCP tools fit into an agentic workflow concept. They explore building in public tactics and monetization, debating Subscription Tier Recommendation, hackathon sponsorship models, and the Pulse MCP tool suite as a business idea that complements open source community growth. Technically, they evaluate architecture decisions with the Circuit Tracing Technique and Observability Concept to improve debugging and memory consistency across agents, and discuss integration patterns for Ai agents and Api Integration. Throughout, practical developer workflows and entrepreneurial trade-offs guide the discussion, balancing rapid iteration with production reliability. They close with a forward-looking call for builders to prioritize memory-first architectures, observability, and public iteration as the foundation for scalable AI startups.