Atlas › 03 Agentic Testing Architectures › Testing Agentic Systems: MCP, A2A, and Agent Evals☰ Read as one page
Testing Agentic Systems: MCP, A2A, and Agent Evals
14.1The Tables Have TurnedEvery previous chapter in this module used agents to test conventional software. This chapter inverts the relationship: the system under…14.2Surface 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. This is a…14.3Surface 2: Testing A2A FlowsA2A (Agent-to-Agent) standardizes how agents from different vendors and runtimes discover and delegate to each other: an agent card…14.4Surface 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…14.5Surface 4: Sandboxing the Agent Under TestIn Guardrails, you constrained your testing agent. When the SUT is an agent, the same thinking applies with the roles reversed -- and…14.6Surface 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: run each scenario…14.7Surface 6: Guardrail TestingThe agent's guardrails -- allowlists, budgets, refusal behavior, injection defenses -- are product requirements, and they need adversarial…14.8The Architect's ChecklistWhen you inherit an agentic system to test, work the surfaces in this order:14.9Key TakeawayTesting agentic systems is classical test discipline aimed at new surfaces: contract tests for MCP tool schemas and A2A agent cards…