Library › Book 3 › Testing Agentic Systems -- MCP, A2A, and Agent Evals
Testing Agentic Systems -- MCP, A2A, and Agent Evals
16.1🔒The Tables Have TurnedEvery previous chapter in this book used agents to test conventional software. This chapter inverts the relationship: the system under test…
16.2🔒Surface 1: Contract Testing MCP Tool IntegrationsAn agent's tools are exposed through MCP (Model Context Protocol) servers, each publishing a list of tools with JSON schemas (Chapter…
16.3🔒Surface 2: Testing A2A FlowsA2A (Agent-to-Agent) standardizes how agents from different vendors and runtimes discover and delegate to each other: an agent card…
16.4🔒Surface 3: Deterministic Replay of Agent TracesYou cannot debug what you cannot reproduce. The core technique is to record the full trace of an agent run -- every model call, every tool…
16.5🔒Surface 4: Sandboxing the Agent Under TestIn Chapter 9, you constrained your testing agent. When the SUT is an agent, the same thinking applies with the roles reversed -- and…
16.6🔒Surface 5: Evaluating Non-Determinism with pass^kA single passing run of an agentic scenario means almost nothing. The standard as of July 2026 is repeat-run evaluation (introduced as…
16.7🔒Surface 6: Guardrail TestingThe agent's guardrails -- allowlists, budgets, refusal behavior, injection defenses -- are product requirements, and they need adversarial…
16.8🔒The Architect's ChecklistWhen you inherit an agentic system to test, work the surfaces in this order:
16.9🔒Career Translation- Built contract-test suite for MCP tool integrations (schema snapshots in git, drift detection in CI), catching 100% of breaking…
16.10🔒Q&AInterview Depth CheckPrompt: Your company's support agent calls five MCP tools. A vendor renames a parameter in one tool's schema over the weekend. Nothing…