Designing Agentic Systems

Designing Agentic Systems

AdvancedTopic5 lessons~4h 5m

Lessons, in order

01 Anatomy of an AgentAn agent is a loop: the model reads the context, decides an action, the harness executes it, the result goes back into the context, repeat until the model says it is finished. The model … Advanced 45 min 02 Orchestration and Durable ExecutionAn agent task can run for an hour, wait a day for a human, and span a hundred model and tool calls. In that time the process running it will be restarted — by a deploy, an autoscaler, a … Advanced 55 min 03 Context Engineering and MemoryAn agent's context grows with every step, and three things go wrong as it does: the window fills, the bill grows quadratically, and the model's attention degrades — it starts to lose track … Advanced 50 min 04 Multi-Agent SystemsSeveral agents are better than one in exactly three situations: the work splits into independent parts that can run in parallel, the work needs more context than one window can hold cleanly, … Advanced 45 min 05 Sandboxes and Tool ExecutionCapable agents write and run code. That code is produced by a model that may have just read an attacker's instructions, so it must be treated as untrusted code and run where it cannot hurt … Advanced 50 min

About this topic

An agent is a model that chooses its own next step, in a loop, until the job is done. That one change — the model decides the control flow — turns a request into a long-running, stateful, expensive process that can act on the world. This topic covers what must be designed around the loop for it to be dependable.

#	Lesson	The question it answers
01	Anatomy of an Agent	What is an agent made of, and when should I not build one?
02	Orchestration and Durable Execution	How does an hour-long task survive a crash, a deploy and a human pause?
03	Context Engineering and Memory	How does a long task stay inside the context window and stay coherent?
04	Multi-Agent Systems	When do several agents beat one, and how do they talk?
05	Sandboxes and Tool Execution	Where does model-written code run, and what contains it?

Read Tools and Protocols and Reliability Patterns first; this topic builds directly on both.