Season 9
Build, evaluate, and harden autonomous multi-agent systems: loop architectures, Model Context Protocol (MCP), and production guardrails.
The agents have a plan. Check it.
What you will cover
Episode 1
The Runaway Ticket
Autonomous agent architectures: the Perception-Reasoning-Action loop (perception of state, tool planning, invocation, observation injection, termination).
Episode 2
The Universal Connector
Model Context Protocol (MCP) architecture: host applications, protocol clients, and decoupled servers communicating over JSON-RPC 2.0.
Episode 3
The Swarm and the Supervisor
Multi-agent system architecture: Supervisor-Worker pattern, intent routing, and compound query decomposition.
Episode 4
The Judge and the Benchmark
Trajectory evaluation: tool-call precision, tool-call recall, argument accuracy, and excess-turn penalty scoring.
Episode 5
The Vault and the Thread
Recent-message buffers: slice a positive-size window, append evicted text and identify unbounded history growth and overflow cases.
Episode 6
The Autonomous Release
Worker topology: task routing and controlled catalog aggregation using fictional tool-server aliases, with host-side authorisation kept explicit.
5 practice projects