The Agent Engineering Bible: 25 Production Agent Builds and 1,000+ Prompts for Multi-Agent AI Systems, MCP and A2A
English | May 27, 2026 | ASIN: B0H34PFW8B | 798 pages | Epub | 558.29 KB
Most AI agent books show you patterns. This one ships agents. Twenty-five complete production agent builds, taught end to end. Every system prompt, every tool schema, every eval harness, every failure mode-copy-pasteable and measured in 2026 dollars. No demoware. No vague abstractions about "AI agents are the future." Production code, production prompts, production cost models. What you get: 25 full agent builds at ~16 pages each - Customer Support, Code Review, Coding Pair, Data Engineer, QA Test Generator, Sales SDR, Marketing Content Studio, Customer Onboarding, Churn Prediction, Research Analyst, Compliance Review, Earnings Analyst, Investment Research, Legal Contract Review, Medical Triage, Recruiting Sourcing, Personalized Tutor, Security Triage, Incident Response, Developer Documentation, Inbox Zero, Calendar Negotiation, Product Spec Writer, Browser Agent, Form Filler 1,000+ production-tested prompts you can adapt today MCP (Model Context Protocol) server design from first principles - tool schemas, error contracts, build vs buy decisions A2A (Agent-to-Agent) protocol with full worked examples - identity, attestation chains, negotiation patterns Eval harnesses you can actually ship - LLM judges, trajectory eval, deploy gating, regression suites Cost models with real 2026 numbers - per-token, per-task, per-outcome frames for every build Memory architecture (ephemeral, episodic, semantic, procedural) with contracts per build Seven-layer safety guardrail stack with prompt-injection defense Human-in-the-loop UX patterns - the draft-don't-decide pattern, interruption tax framework Observability, cost monitoring, regression-driven improvement The frontier - agent commerce (x402, AP2), Computer Use depth, on-device agents, multimodal, self-improving agents, 18-month forecast Reference appendices - MCP server catalog (60+ servers), framework comparison (11 frameworks), cost reference tables, model selection matrix, cross-reference index, glossary Who it's for: Engineers shipping production AI agents. Tech leads making 12-18 month architecture bets. Product owners who need to know what's possible at what cost. CTOs deciding build vs buy. Anyone tired of "AI agents are transformative" content and ready for "here's the system prompt, here's the eval, here's the cost per task." Specific models, specific protocols, specific numbers. Claude Sonnet 4.6 and Opus 4.7, GPT-5 and the o-series, Gemini 3 Pro, Llama 4 70B, plus the full MCP server ecosystem, the x402 payment protocol, the A2A protocol drafts, DSPy, LangGraph, Mastra, Inngest, Temporal, and the 60+ MCP servers currently in production use. Pricing as of May 2026, with forecasts grounded in measurable trends. The standalone companion to the four-volume AI Engineering Bible series. Read it on its own; read it next to Volumes I-IV; either way, ship better agents. 720 pages. 25 builds. 1,000+ prompts. No filler.
Links are Interchangeable - Single Extraction
