"Agent Communication Protocol (ACP): Designing Secure Agent-to-Agent APIs and Workflows"
Agents are quickly becoming the new integration surface—and the moment you let them call each other across services, teams, or organizations, the "it's just an API" assumptions stop working. This book is written for senior engineers, security architects, and platform teams building agent-to-agent systems that must survive real production traffic, long-running workflows, and hostile inputs. It treats ACP as an interoperability contract and shows how to design systems that remain composable without becoming dangerously permissive.
You'll learn how to define protocol scope and roles (principals, agents, domains), then design transport contracts that handle retries, streaming, cancellation, and backpressure without creating "exactly-once" myths. The book goes deep on manifests, discovery, and handshake—treating capability descriptions as policy inputs, not marketing—followed by a rigorous message model with canonicalization, typed message parts, and downstream-safe rendering. It then formalizes runs, sessions, and await/resume as securable state machines, and builds a protocol-centric security program spanning identity, authentication, authorization, confidentiality, integrity, and replay defense. Finally, you'll apply these primitives to delegation, fan-out/fan-in orchestration, observability, incident response, and conformance testing.
Expect a dense, implementation-oriented treatment with explicit failure modes, anti-patterns, and migration-aware compatibility guidance. Familiarity wi
