September 2026
Intermediate
774 pages
19h 47m
English
This chapter moves from production to persistence. Once agents can run in production, the next architectural question is what they remember, what they retain as evidence, and how that state survives restarts, upgrades, and audits without becoming an uncontrolled liability.
By now, we have deployed the agent. The Cloud Run containers, Agent Development Kit (ADK) runners, and MCP tool integrations from Chapter 8 turn our reference case study, ThesisBot, into a viable system that runs.
The next test is what happens after the request is over. Can the agent prove what it decided, why it decided it, and when? There is always an audit in one shape or another, and that means we need to talk about memory management, ...
Read now
Unlock full access