Collections

Best Chain of Thought Episodes on Multi-Agent Systems

Many agents, one system — orchestration patterns, protocols, and the overhead they add.

One agent is a software problem. Several agents talking to each other is a distributed-systems problem: partial failure and handoffs, running on models that don’t behave the same way twice. All four of these guests hit that in production rather than on a whiteboard, and they disagree with each other about where the coordination belongs.

  1. 1 EP 33 40 min Transcript Mastering Multi-Agent Systems | MongoDB’s Mikiko Chandrasekhar Mikiko Chandrasekhar, MongoDB Mikiko Chandrasekhar on treating agents as software products and applying real engineering discipline to them — the thing that keeps a multi-agent system reliable when nothing inside it is deterministic.
  2. 2 EP 29 50 min Transcript The Emerging AI Agent Stack | CrewAI’s João Moura João Moura, CrewAI CrewAI’s João Moura maps the orchestration layer everyone is racing to standardize, and is blunt about how many more failure modes you inherit the moment there is more than one agent.
  3. 3 EP 21 39 min Transcript Why Enterprises Need a Different Approach to AI Agents | Lyzr’s Siva Surendira Siva Surendira, Lyzr Lyzr’s Siva Surendira walks the actual orchestration shapes — managerial, DAG, hybrid — and where each one runs into an enterprise’s integration and data-readiness limits.
  4. 4 EP 31 41 min Transcript Architecting Reliable Agentic AI | Cisco’s Giovanna Carofiglio on the AGNTCY Collective Giovanna Carofiglio, Cisco Cisco’s Giovanna Carofiglio on AGNTCY and the Internet of Agents: agent discovery, the transport layer under agent-to-agent traffic, and why MCP and A2A are solving different problems.

More collections

All collections →

New here? Start with five → Browse by topic →