AI Agents in Production: AgentOps — Deploying, Scaling, and Operating Agents at Work. Durability, cost, monitoring, versioning, identity, scaling, and incident response, built into one platform - Couverture souple

MOBILUCK - CODE247.AI, VU TRI CONG

Couverture souple

ISBN 13 : 9798171351267

Editeur : Independently published, 2026

Synopsis

The agent works on your laptop. That is exactly where most agent projects quietly die.

A working prototype is not a production system. In production, requests arrive concurrently at all hours, processes crash and restart, models drift under stable versions, costs accrue per call, and one bad input must never take down the next ninety-nine callers. This book is about everything around the agent's mind — the operational discipline the industry calls AgentOps — that turns a program you run into a service a business can depend on. Its running system is a platform that operates the prior books' agents, and across ten chapters it builds that platform one discipline at a time.

Through ten chapters, each moving from a reproduced production failure to tested code, you will build:

• The agent as a service — a contract, health check, version stamp, and error isolation, the shape everything else attaches to


• Durable execution — state outside the process, checkpoints, leases, and idempotency, so work survives a crash or a deploy


• Cost engineering — token metering with attribution, caching, and small-vs-large model routing, the biggest lever there is


• Monitoring with SLOs over rolling windows — catching the quiet drift, not just the loud errors


• Versioning with canary and one-call rollback — change proven in production before it owns production


• The data lifecycle — refresh, hygiene, retention, and tombstone deletes, because an agent is only as current as its knowledge


• Least-privilege identity — the agent as a principal with its own scoped credential, deny-by-default, every attempt audited


• Scaling multi-agent teams to the coordination sweet spot — where more agents makes the team slower, and why


• Incident response — mitigate before you understand, resolve blamelessly, feed the fix back


• The platform — all nine disciplines composed into one control plane that operates agents as services

The governing idea, held for ten chapters: the platform, not the agent, is the durable asset — models, agents, and tasks churn, but the invariants of operating an agent do not, so the operational surface outlives every agent it runs. Each chapter adds a research lineage (the twelve-factor app and microservices to Google SRE, Amdahl and Brooks to the Kubernetes control plane), a failure catalog, an applied deep-dive, and exercises. The companion repo has one dependency, twenty-two tests, and runs entirely offline.

Who it's for: platform, DevOps, and senior engineers putting agents into production, who have built, dissected, and hardened an agent (Books 3–5 or equivalent).

The AI Agent Series takes readers from first contact to professional mastery across seven volumes. This is Book 6: agents in production.

Détails de l'édition

• Éditeur: Independently published
• Date d'édition: 2026
• Langue: anglais
• ISBN 13: 9798171351267
• Reliure: Broché
• Nombre de pages: 163
• Coordonnées du fabricant: Manufactured by Amazon on behalf of the author; https://www.amazon.fr/hz/contact-us; ; c/o Amazon Media EU S.à.r.l., 38 Avenue John F. Kennedy; Luxembourg; L-1855; Luxembourg