The AI Agent Field Manual

₹1,900

Everything you need to take an AI agent from prototype to production — with the receipts. The AI Agent Field Manual is a vendor-neutral, citation-backed guide to building, orchestrating, and operating AI agents at a production standard. Where a quick-start guide gets you moving, this book is built for engineers who want the full picture: the architectural reasoning, the failure research, and the infrastructure patterns that separate a demo from a system you'd trust in production. Every chapter pairs a clear explanation with a Key Points summary and a practical example — real Python snippets and architecture diagrams you can map directly onto systems you already run. What's inside: What an AI agent actually is, and the ReAct loop that powers it Function calling and the Model Context Protocol (MCP) — with a full JSON Schema example All five workflow orchestration patterns (chaining, routing, parallelization, orchestrator–workers, evaluator–optimizer) plus when to reach for a full autonomous agent Memory, RAG, context engineering, and checkpointing — including the 2025–2026 shift toward just-in-time retrieval Multi-agent communication (shared state, handoffs, A2A) — and a candid look at the research on why multi-agent systems fail (the MAST study, Anthropic's own numbers, Cognition's counterpoint) Production infrastructure: durable execution, idempotency, sandboxing, and prompt caching Observability and evaluation: tracing, LLM-as-a-judge, and the offline/online eval lifecycle Security: prompt injection, the OWASP Agentic Top 10, and defense-in-depth you can actually implement A full end-to-end worked example: a support-ticket triage system tying every chapter together, with complete code A pitfalls checklist, a step-by-step learning path, and a glossary of every term used Who this is for: Software and backend engineers who want to build production AI agents on solid architectural footing — not just follow a tutorial, but understand why each pattern exists and what the research says about when it breaks.