Topic 05 Practice Test 3 covers Agentic AI, Tool Calling, Human Review, Strands, Lambda, ECS, and MCP in Domain 2 (Task 2.1). For broader exam preparation, review the AIP-C01 Exam Dumps. Every option includes focused technical reasoning tied to the tested decision.
Question 1
A case-management agent needs short-term conversational memory for the active case but should not embed durable business state solely in the model prompt. Which option most directly addresses the requirement? Choose ONE.
- Application-managed durable state
- MCP for agent-tool interoperability
- Custom result aggregation
- Strands Agents with explicit session state
Correct Answer: D
Correct Answer
Answer D is correct because Strands Agents with explicit session state — for the described workload, uses the Strands agent framework, for the stated need, while preserving the state needed, for the stated need, across turns or workflow steps. Best suited when an agent must retain task/session, for this decision, context while coordinating tools. Deciding factor: manage active agent session context while keeping business state separate. Strands Agents with explicit session state fits this case.
Incorrect Answers
Answer A is incorrect because Application-managed durable state — under these constraints, stores durable workflow/session state outside, at this layer, the model so progress survives, at this layer, process restarts. Choose it when a long-running agent task must, under these constraints, resume reliably after infrastructure interruption. Requirement: manage active agent session context while keeping business state separate. Application-managed durable state is not the direct fit.
Answer B is incorrect because MCP for agent-tool interoperability — in the current architecture, uses the Model Context Protocol, in the described flow, to expose tools through a, in the described flow, consistent agent-to-tool interface. Applies where tools should be reachable through, in this case, a standardized protocol instead of, in this case, bespoke integrations. Target outcome: manage active agent session context while keeping business state separate. MCP for agent-tool interoperability misses that outcome.
Answer C is incorrect because Custom result aggregation — within this implementation, combines outputs from multiple model, for this decision, or agent workers using explicit, for this decision, application logic. Strong fit when several parallel specialist results must, within this design, be reconciled into one answer. Operational need: manage active agent session context while keeping business state separate. Custom result aggregation is secondary here.
Question 2
A complex procurement task should delegate vendor research, policy checking, and contract review to different specialists. Which design best satisfies this requirement? Choose ONE.
- MCP client library
- AWS Agent Squad multi-agent coordination
- Custom result aggregation
- Specialist-model routing
Correct Answer: B
Correct Answer
Answer B is correct because AWS Agent Squad multi-agent coordination — within this implementation, coordinates multiple specialized agents so, for this decision, each can contribute a bounded, for this decision, capability to a larger task. Strong fit when a workflow benefits from several, within this design, specialist agents rather than one, within this design, monolithic agent. Operational need: coordinate distinct agent roles for one composite task. AWS Agent Squad multi-agent coordination is appropriate here.
Incorrect Answers
Answer A is incorrect because MCP client library — in the current architecture, gives agents a consistent client-side, in the described flow, mechanism for discovering and invoking, in the described flow, MCP-exposed tools. Relevant when multiple MCP servers should be, in this case, accessed through a common client, in this case, pattern. The scenario tests coordinate distinct agent roles for one composite task. MCP client library addresses different behavior.
Answer C is incorrect because Custom result aggregation — in this scenario, combines outputs from multiple model, in this case, or agent workers using explicit, in this case, application logic. Appropriate if several parallel specialist results must, at this layer, be reconciled into one answer. The stem requires coordinate distinct agent roles for one composite task. Custom result aggregation serves another purpose.
Answer D is incorrect because Specialist-model routing — in this scenario, routes a request to a, in this case, model selected for the task’s, in this case, capability, latency, or quality needs. Appropriate if different task classes benefit from, at this layer, different model strengths. The stem requires coordinate distinct agent roles for one composite task. Specialist-model routing serves another purpose.
Question 3
A company wants external tool teams to publish capabilities through a standard interface that different agents can consume. What is the most appropriate implementation choice? Choose ONE.
- API Gateway feedback endpoint
- Standardized function schema
- RETURN_CONTROL action group
- MCP for agent-tool interoperability
Correct Answer: D
Correct Answer
Answer D is correct because MCP for agent-tool interoperability — for this operational need, uses the Model Context Protocol, under these constraints, to expose tools through a, under these constraints, consistent agent-to-tool interface. Architects use it when tools should be reachable through, for this workload, a standardized protocol instead of, for this workload, bespoke integrations. Key need: standardize publication and invocation of tools across agent implementations. MCP for agent-tool interoperability aligns with it.
Incorrect Answers
Answer A is incorrect because API Gateway feedback endpoint — within this implementation, provides an HTTP interface through, for this decision, which applications can submit structured, for this decision, human feedback or approval data. Strong fit when the workflow needs an external, within this design, feedback collection endpoint. Operational need: standardize publication and invocation of tools across agent implementations. API Gateway feedback endpoint is secondary here.
Answer B is incorrect because Standardized function schema — under these constraints, defines tool names, parameters, types,, at this layer, and required fields consistently so, at this layer, the model can call tools, at this layer, reliably. Makes sense when tool invocation errors are caused, under these constraints, by ambiguous or inconsistent function, under these constraints, definitions. The case depends on standardize publication and invocation of tools across agent implementations. Standardized function schema handles another concern.
Answer C is incorrect because RETURN_CONTROL hands an agent’s selected action and parameters back to the calling application for external execution. It does not define a portable interface through which independent tool teams publish capabilities for different agents. MCP addresses the cross-agent tool publication and invocation standardization required by the scenario.
Question 4
A high-risk autonomous workflow needs explicit branches for approval, retry, rollback, and final completion. Which approach should the developer use? Choose ONE.
- Application-managed durable state
- Step Functions orchestration
- Compensating workflow
- Human approval in Step Functions
Correct Answer: B
Correct Answer
Answer B is correct because Step Functions orchestration — under these constraints, models multi-step workflow control, branching,, at this layer, retries, approvals, and stopping conditions, at this layer, explicitly. Choose it when an AI workflow needs deterministic, under these constraints, orchestration around model or tool, under these constraints, calls. Requirement: represent complex control flow as auditable workflow states. Step Functions orchestration addresses it precisely.
Incorrect Answers
Answer A is incorrect because Application-managed durable state — for the described workload, stores durable workflow/session state outside, for the stated need, the model so progress survives, for the stated need, process restarts. Best suited when a long-running agent task must, for this decision, resume reliably after infrastructure interruption. Deciding factor: represent complex control flow as auditable workflow states. Application-managed durable state addresses something else.
Answer C is incorrect because Compensating workflow — at this decision point, defines explicit reversal or remediation, within this design, steps when a multi-step agentic, within this design, transaction partially succeeds. Typical use: a later tool failure must, in the described flow, safely unwind earlier side effects. Goal: represent complex control flow as auditable workflow states. Compensating workflow does not provide it.
Answer D is incorrect because Human approval in Step Functions — in the current architecture, pauses an orchestrated workflow for, in the described flow, an explicit human review or, in the described flow, approval before continuing. Applies where high-impact actions require human judgment, in this case, before execution. Target outcome: represent complex control flow as auditable workflow states. Human approval in Step Functions misses that outcome.
Question 5
A Strands agent calls an inventory tool that can occasionally return a transient 503 error. The application should retry that specific tool invocation once immediately after the failed result is observed, without wrapping the agent in a Step Functions workflow. Which implementation best fits? Choose ONE.
- Strands BeforeToolCall hook that edits request arguments
- Lambda timeout guard
- Strands AfterToolCall hook that requests a retry
- Step Functions retry with backoff
Correct Answer: C
Correct Answer
Answer C is correct because A Strands AfterToolCall hook runs after the tool invocation has completed, including when the call returns an error. The hook can inspect that failed result and request a retry of the tool invocation, so it matches the requirement to retry once inside the agent lifecycle after the transient 503 is observed.
Incorrect Answers
Answer A is incorrect because A BeforeToolCall hook runs before the tool executes and is useful for inspecting, modifying, replacing, or canceling the pending call. It cannot base its decision on a 503 result that has not happened yet, so it is the wrong lifecycle point for the required post-execution retry behavior.
Answer B is incorrect because A Lambda timeout guard limits how long a function invocation may run. It does not react to a completed tool result by asking the Strands tool executor to invoke that same tool again. The requirement is a targeted retry after a transient failure, not a maximum execution-duration control.
Answer D is incorrect because Step Functions can retry failed workflow tasks with policies such as backoff, but the scenario explicitly keeps the behavior inside the existing Strands agent lifecycle and avoids adding a state-machine wrapper. An AfterToolCall hook provides the narrower mechanism for reacting to this tool result and requesting the retry directly.
Question 6
A model can request database reads, but the execution role must permit access only to one table and no write operations. Which option most directly addresses the requirement? Choose ONE.
- IAM least-privilege boundary
- Per-tool IAM role
- Tool allowlist
- Lambda parameter validation
Correct Answer: A
Correct Answer
Answer A is correct because IAM least-privilege boundary — in the current architecture, limits the agent or tool, in the described flow, role to only the AWS, in the described flow, actions and resources required for, in the described flow, the task. Applies where resource access must be constrained, in this case, independently of model behavior. Target outcome: constrain the AWS resource actions available to the agent tool. IAM least-privilege boundary supplies it directly.
Incorrect Answers
Answer B is incorrect because Per-tool IAM role — at this decision point, assigns narrowly scoped permissions to, within this design, a specific tool integration rather, within this design, than sharing a broad execution, within this design, role. Typical use: different tools have distinct resource-access requirements. Goal: constrain the AWS resource actions available to the agent tool. Per-tool IAM role does not provide it.
Answer C is incorrect because Tool allowlist — in this scenario, restricts the agent to an, in this case, approved set of callable operations, in this case, instead of exposing every available, in this case, action. Appropriate if the risk boundary is which, at this layer, tools the model may invoke. The stem requires constrain the AWS resource actions available to the agent tool. Tool allowlist serves another purpose.
Answer D is incorrect because Lambda parameter validation — within this implementation, validates and normalizes tool inputs, for this decision, before business logic or downstream, for this decision, API execution. Strong fit when tool calls need defensive checks, within this design, before side effects occur. Operational need: constrain the AWS resource actions available to the agent tool. Lambda parameter validation is secondary here.
Question 7
A shipping carrier API has been failing for every request; after several failures the application should stop trying until a cooldown expires. Which design best satisfies this requirement? Choose ONE.
- Circuit breaker around failing tool
- Idempotent tool operation
- Lambda timeout guard
- Step Functions retry with backoff
Correct Answer: A
Correct Answer
Answer A is correct because A circuit breaker tracks repeated failures and, after the configured threshold, stops sending requests until a cooldown or recovery condition occurs. That prevents every new agent interaction from continuing to hit the unhealthy carrier API. The behavior exactly matches the requirement to pause calls after persistent failures.
Incorrect Answers
Answer B is incorrect because Idempotent tool operation — for this operational need, uses an idempotency token or, under these constraints, state check so a retried, under these constraints, action does not create duplicate, under these constraints, side effects. Architects use it when workflow retries could otherwise repeat, for this workload, a charge, ticket, or update. Key need: protect the system from a persistently unhealthy dependency. Idempotent tool operation does not satisfy it.
Answer C is incorrect because Lambda timeout guard — at this decision point, enforces bounded execution around a, within this design, tool or model-supporting function so, within this design, stalled work cannot run indefinitely. Use it where a lightweight serverless action needs, in the described flow, a clear execution timeout. The design requires protect the system from a persistently unhealthy dependency. Lambda timeout guard is not the direct choice.
Answer D is incorrect because Retry with backoff can make a few failed calls less aggressive, but each new workflow can still continue attempting the same unhealthy carrier endpoint. The requirement is to suppress calls after a failure threshold until a cooldown expires. A circuit breaker provides that shared open-state behavior across attempts.
Question 8
An image-understanding request and a long-form planning request should not be forced through the same foundation model. What is the most appropriate implementation choice? Choose ONE.
- Custom result aggregation
- AWS Agent Squad multi-agent coordination
- Model-selection framework
- Specialist-model routing
Correct Answer: D
Correct Answer
Answer D is correct because Specialist-model routing — for this design, routes a request to a, for this workload, model selected for the task’s, for this workload, capability, latency, or quality needs. Supports designs where different task classes benefit from, for the stated need, different model strengths. The workload needs choose models by task capability rather than use one universal model. Specialist-model routing maps to it.
Incorrect Answers
Answer A is incorrect because Custom result aggregation — for this design, combines outputs from multiple model, for this workload, or agent workers using explicit, for this workload, application logic. Supports designs where several parallel specialist results must, for the stated need, be reconciled into one answer. The workload needs choose models by task capability rather than use one universal model. Custom result aggregation maps elsewhere.
Answer B is incorrect because Multi-agent coordination assigns separate responsibilities to multiple agents and combines their contributions. The problem here is selecting a foundation model based on the modality and reasoning needs of each request, not coordinating several agents. Specialist-model routing can send image work to a vision-capable model and planning work elsewhere.
Answer C is incorrect because Model-selection framework — for this operational need, applies policy or scoring logic, under these constraints, to choose among candidate models, under these constraints, at runtime. Architects use it when the application must choose a, for this workload, model based on measurable request, for this workload, characteristics. Key need: choose models by task capability rather than use one universal model. Model-selection framework does not satisfy it.
Question 9
A coordinator receives proposals from multiple agents and must apply deterministic tie-breaking and policy rules before returning one plan. Which approach should the developer use? Choose ONE.
- AWS Agent Squad multi-agent coordination
- Model-selection framework
- Specialist-model routing
- Custom result aggregation
Correct Answer: D
Correct Answer
Answer D is correct because Custom result aggregation — at this decision point, combines outputs from multiple model, within this design, or agent workers using explicit, within this design, application logic. Use it where several parallel specialist results must, in the described flow, be reconciled into one answer. The design requires merge multi-agent outputs under explicit application policy. Custom result aggregation is the direct choice.
Incorrect Answers
Answer A is incorrect because AWS Agent Squad multi-agent coordination — for this design, coordinates multiple specialized agents so, for this workload, each can contribute a bounded, for this workload, capability to a larger task. Supports designs where a workflow benefits from several, for the stated need, specialist agents rather than one, for the stated need, monolithic agent. The workload needs merge multi-agent outputs under explicit application policy. AWS Agent Squad multi-agent coordination maps elsewhere.
Answer B is incorrect because Model-selection framework — within this implementation, applies policy or scoring logic, for this decision, to choose among candidate models, for this decision, at runtime. Strong fit when the application must choose a, within this design, model based on measurable request, within this design, characteristics. Operational need: merge multi-agent outputs under explicit application policy. Model-selection framework is secondary here.
Answer C is incorrect because Specialist-model routing — at this decision point, routes a request to a, within this design, model selected for the task’s, within this design, capability, latency, or quality needs. Use it where different task classes benefit from, in the described flow, different model strengths. The design requires merge multi-agent outputs under explicit application policy. Specialist-model routing is not the direct choice.
Question 10
The application has measured quality, latency, and cost scores for several models and must choose one per request using those scores. What should the team implement? Choose ONE.
- Custom result aggregation
- Specialist-model routing
- Model-selection framework
- Standardized function schema
Correct Answer: C
Correct Answer
Answer C is correct because Model-selection framework — in this scenario, applies policy or scoring logic, in this case, to choose among candidate models, in this case, at runtime. Appropriate if the application must choose a, at this layer, model based on measurable request, at this layer, characteristics. The stem requires make runtime model selection from measurable policy criteria. Model-selection framework meets that requirement.
Incorrect Answers
Answer A is incorrect because Custom result aggregation — under these constraints, combines outputs from multiple model, at this layer, or agent workers using explicit, at this layer, application logic. Makes sense when several parallel specialist results must, under these constraints, be reconciled into one answer. The case depends on make runtime model selection from measurable policy criteria. Custom result aggregation handles another concern.
Answer B is incorrect because Specialist-model routing — under these constraints, routes a request to a, at this layer, model selected for the task’s, at this layer, capability, latency, or quality needs. Makes sense when different task classes benefit from, under these constraints, different model strengths. The case depends on make runtime model selection from measurable policy criteria. Specialist-model routing handles another concern.
Answer D is incorrect because A standardized function schema improves the reliability of tool calls by defining names, parameters, types, and required fields. It does not evaluate foundation models against measured quality, latency, and cost scores. A model-selection framework can apply those metrics as policy and choose the appropriate model for each request.
Question 11
A clinical-support agent can prepare a draft note but a licensed reviewer must approve it before it enters the patient record. Which option most directly addresses the requirement? Choose ONE.
- RETURN_CONTROL action group
- API Gateway feedback endpoint
- Tool allowlist
- Human approval in Step Functions
Correct Answer: D
Correct Answer
Answer D is correct because Human approval in Step Functions — for this operational need, pauses an orchestrated workflow for, under these constraints, an explicit human review or, under these constraints, approval before continuing. Architects use it when high-impact actions require human judgment, for this workload, before execution. Key need: ensure a human approval step precedes a sensitive system-of-record update. Human approval in Step Functions aligns with it.
Incorrect Answers
Answer A is incorrect because RETURN_CONTROL gives the caller the agent’s proposed action and parameters so the application can execute the action itself. That handoff does not inherently require a licensed reviewer to approve the draft before the patient record changes. A human-approval workflow provides the explicit pause and authorization gate the scenario requires.
Answer B is incorrect because API Gateway feedback endpoint — in this scenario, provides an HTTP interface through, in this case, which applications can submit structured, in this case, human feedback or approval data. Appropriate if the workflow needs an external, at this layer, feedback collection endpoint. The stem requires ensure a human approval step precedes a sensitive system-of-record update. API Gateway feedback endpoint serves another purpose.
Answer C is incorrect because Tool allowlist — for this design, restricts the agent to an, for this workload, approved set of callable operations, for this workload, instead of exposing every available, for this workload, action. Supports designs where the risk boundary is which, for the stated need, tools the model may invoke. The workload needs ensure a human approval step precedes a sensitive system-of-record update. Tool allowlist maps elsewhere.
Question 12
A reviewer portal needs an authenticated web API for sending corrections and approval decisions to an agent workflow. Which design best satisfies this requirement? Choose ONE.
- API Gateway feedback endpoint
- Lambda action implementation
- Human approval in Step Functions
- MCP client library
Correct Answer: A
Correct Answer
Answer A is correct because API Gateway feedback endpoint — for this design, provides an HTTP interface through, for this workload, which applications can submit structured, for this workload, human feedback or approval data. Supports designs where the workflow needs an external, for the stated need, feedback collection endpoint. The workload needs accept external human feedback through an HTTP API. API Gateway feedback endpoint maps to it.
Incorrect Answers
Answer B is incorrect because Lambda action implementation — in this scenario, runs bounded custom business logic, in this case, for an agent tool or, in this case, action group in a serverless, in this case, function. Appropriate if a tool action is lightweight,, at this layer, stateless, and fits Lambda execution, at this layer, constraints. The stem requires accept external human feedback through an HTTP API. Lambda action implementation serves another purpose.
Answer C is incorrect because Human approval in Step Functions — within this implementation, pauses an orchestrated workflow for, for this decision, an explicit human review or, for this decision, approval before continuing. Strong fit when high-impact actions require human judgment, within this design, before execution. Operational need: accept external human feedback through an HTTP API. Human approval in Step Functions is secondary here.
Answer D is incorrect because MCP client library — for this operational need, gives agents a consistent client-side, under these constraints, mechanism for discovering and invoking, under these constraints, MCP-exposed tools. Helps when multiple MCP servers should be, for this workload, accessed through a common client, for this workload, pattern. The required capability is accept external human feedback through an HTTP API. MCP client library does not supply it.
Question 13
A Bedrock agent action performs a small tax-rate lookup and formatting task that completes quickly and needs no server process. What is the most appropriate implementation choice? Choose ONE.
- Stateless MCP server on Lambda
- MCP server on Amazon ECS
- Lambda action implementation
- Application-managed durable state
Correct Answer: C
Correct Answer
Answer C is correct because Lambda action implementation — for this design, runs bounded custom business logic, for this workload, for an agent tool or, for this workload, action group in a serverless, for this workload, function. Supports designs where a tool action is lightweight,, for the stated need, stateless, and fits Lambda execution, for the stated need, constraints. The workload needs execute a simple stateless action with serverless compute. Lambda action implementation maps to it.
Incorrect Answers
Answer A is incorrect because Stateless MCP server on Lambda — for the described workload, hosts lightweight, stateless MCP tool, for the stated need, access in Lambda with minimal, for the stated need, server management. Best suited when the MCP workload is short-lived,, for this decision, stateless, and event-driven. Deciding factor: execute a simple stateless action with serverless compute. Stateless MCP server on Lambda addresses something else.
Answer B is incorrect because MCP server on Amazon ECS — for this operational need, hosts an MCP server in, under these constraints, containers suitable for longer-lived, more, under these constraints, complex, or dependency-heavy tool services. Architects use it when the tool server needs persistent, for this workload, process behavior or containerized dependencies. Key need: execute a simple stateless action with serverless compute. MCP server on Amazon ECS does not satisfy it.
Answer D is incorrect because Application-managed durable state — in the current architecture, stores durable workflow/session state outside, in the described flow, the model so progress survives, in the described flow, process restarts. Applies where a long-running agent task must, in this case, resume reliably after infrastructure interruption. Target outcome: execute a simple stateless action with serverless compute. Application-managed durable state misses that outcome.
Question 14
A caller must use the agent’s chosen operation and arguments to invoke a mainframe connector that only the caller can reach. Which approach should the developer use? Choose ONE.
- MCP client library
- RETURN_CONTROL action group
- API Gateway feedback endpoint
- Lambda action implementation
Correct Answer: B
Correct Answer
Answer B is correct because RETURN_CONTROL returns the chosen operation and its parameters to the calling application rather than executing fulfillment inside the Bedrock action group. Because only the caller can reach the mainframe connector, this lets the application invoke that connector through its existing network path and then send the result back to the agent.
Incorrect Answers
Answer A is incorrect because MCP client library — within this implementation, gives agents a consistent client-side, for this decision, mechanism for discovering and invoking, for this decision, MCP-exposed tools. Designed for cases where multiple MCP servers should be, within this design, accessed through a common client, within this design, pattern. The key constraint is return the action parameters instead of executing them inside the agent service. MCP client library does not satisfy it.
Answer C is incorrect because API Gateway feedback endpoint — at this decision point, provides an HTTP interface through, within this design, which applications can submit structured, within this design, human feedback or approval data. Use it where the workflow needs an external, in the described flow, feedback collection endpoint. The design requires return the action parameters instead of executing them inside the agent service. API Gateway feedback endpoint is not the direct choice.
Answer D is incorrect because A Lambda fulfillment function would execute the action from AWS infrastructure instead of handing it to the caller. The scenario states that only the calling application can reach the mainframe connector, so moving execution to Lambda does not satisfy the connectivity constraint. RETURN_CONTROL preserves caller-side execution.
Question 15
Tools from several teams should expose consistent names, typed parameters, and required fields so agents can call them predictably. What should the team implement? Choose ONE.
- Lambda parameter validation
- Standardized function schema
- MCP client library
- Tool allowlist
Correct Answer: B
Correct Answer
Answer B is correct because Standardized function schema — in the current architecture, defines tool names, parameters, types,, in the described flow, and required fields consistently so, in the described flow, the model can call tools, in the described flow, reliably. Relevant when tool invocation errors are caused, in this case, by ambiguous or inconsistent function, in this case, definitions. The scenario tests define predictable machine-readable tool contracts. Standardized function schema addresses that behavior.
Incorrect Answers
Answer A is incorrect because Lambda parameter validation — in this scenario, validates and normalizes tool inputs, in this case, before business logic or downstream, in this case, API execution. Appropriate if tool calls need defensive checks, at this layer, before side effects occur. The stem requires define predictable machine-readable tool contracts. Lambda parameter validation serves another purpose.
Answer C is incorrect because MCP client library — in this scenario, gives agents a consistent client-side, in this case, mechanism for discovering and invoking, in this case, MCP-exposed tools. Useful when multiple MCP servers should be, at this layer, accessed through a common client, at this layer, pattern. Need: define predictable machine-readable tool contracts. MCP client library solves another problem.
Answer D is incorrect because Tool allowlist — at this decision point, restricts the agent to an, within this design, approved set of callable operations, within this design, instead of exposing every available, within this design, action. Use it where the risk boundary is which, in the described flow, tools the model may invoke. The design requires define predictable machine-readable tool contracts. Tool allowlist is not the direct choice.
Question 16
A booking tool receives a passenger count and travel date from the model. Values outside the business rules must fail before booking. Which option most directly addresses the requirement? Choose ONE.
- IAM least-privilege boundary
- Idempotent tool operation
- Lambda parameter validation
- Standardized function schema
Correct Answer: C
Correct Answer
Answer C is correct because Lambda parameter validation — for this design, validates and normalizes tool inputs, for this workload, before business logic or downstream, for this workload, API execution. Supports designs where tool calls need defensive checks, for the stated need, before side effects occur. The workload needs enforce business input validation before an irreversible call. Lambda parameter validation maps to it.
Incorrect Answers
Answer A is incorrect because IAM least-privilege boundary — for this operational need, limits the agent or tool, under these constraints, role to only the AWS, under these constraints, actions and resources required for, under these constraints, the task. Architects use it when resource access must be constrained, for this workload, independently of model behavior. Key need: enforce business input validation before an irreversible call. IAM least-privilege boundary does not satisfy it.
Answer B is incorrect because Idempotent tool operation — within this implementation, uses an idempotency token or, for this decision, state check so a retried, for this decision, action does not create duplicate, for this decision, side effects. Strong fit when workflow retries could otherwise repeat, within this design, a charge, ticket, or update. Operational need: enforce business input validation before an irreversible call. Idempotent tool operation is secondary here.
Answer D is incorrect because Standardized function schema — for this operational need, defines tool names, parameters, types,, under these constraints, and required fields consistently so, under these constraints, the model can call tools, under these constraints, reliably. Helps when tool invocation errors are caused, for this workload, by ambiguous or inconsistent function, for this workload, definitions. The required capability is enforce business input validation before an irreversible call. Standardized function schema does not supply it.
Question 17
A small stateless MCP service exposes two lookup tools and has highly bursty demand with long idle periods. Which design best satisfies this requirement? Choose ONE.
- Stateless MCP server on Lambda
- Lambda action implementation
- MCP server on Amazon ECS
- MCP client library
Correct Answer: A
Correct Answer
Answer A is correct because Stateless MCP server on Lambda — in the current architecture, hosts lightweight, stateless MCP tool, in the described flow, access in Lambda with minimal, in the described flow, server management. Applies where the MCP workload is short-lived,, in this case, stateless, and event-driven. Target outcome: host bursty lightweight MCP tools without maintaining servers. Stateless MCP server on Lambda supplies it directly.
Incorrect Answers
Answer B is incorrect because Lambda action implementation — under these constraints, runs bounded custom business logic, at this layer, for an agent tool or, at this layer, action group in a serverless, at this layer, function. Makes sense when a tool action is lightweight,, under these constraints, stateless, and fits Lambda execution, under these constraints, constraints. The case depends on host bursty lightweight MCP tools without maintaining servers. Lambda action implementation handles another concern.
Answer C is incorrect because MCP server on Amazon ECS — within this implementation, hosts an MCP server in, for this decision, containers suitable for longer-lived, more, for this decision, complex, or dependency-heavy tool services. Strong fit when the tool server needs persistent, within this design, process behavior or containerized dependencies. Operational need: host bursty lightweight MCP tools without maintaining servers. MCP server on Amazon ECS is secondary here.
Answer D is incorrect because MCP client library — for this design, gives agents a consistent client-side, for this workload, mechanism for discovering and invoking, for this workload, MCP-exposed tools. Fits when multiple MCP servers should be, for the stated need, accessed through a common client, for the stated need, pattern. Constraint: host bursty lightweight MCP tools without maintaining servers. MCP client library targets another layer.
Question 18
A Strands agent calls a billing tool, but application code must inspect and normalize the tool arguments immediately before the tool executes. The team wants this behavior inside the agent lifecycle without changing the tool schema or adding a separate workflow. Which implementation best fits? Choose ONE.
- Strands BeforeToolCall hook
- MCP client library
- Lambda parameter validation
- Step Functions orchestration
Correct Answer: A
Correct Answer
Answer A is correct because A Strands BeforeToolCall hook runs after tool selection but before execution, so application code can inspect or rewrite the pending tool request in the agent lifecycle. That directly fits a requirement to normalize billing arguments before side effects while leaving the tool contract unchanged and avoiding a separate orchestration service.
Incorrect Answers
Answer B is incorrect because An MCP client library gives an agent a consistent way to discover and invoke tools exposed by MCP servers. It addresses protocol access and client interoperability, not a pre-execution callback within a Strands agent that can inspect or modify the selected tool request before the tool runs.
Answer C is incorrect because Lambda parameter validation is useful when the tool implementation itself runs in Lambda and must reject invalid inputs before performing business logic. It occurs inside the invoked function, not as an agent-lifecycle interception point before the selected tool executes, so it does not satisfy the requested placement of the control.
Answer D is incorrect because Step Functions is appropriate for explicit multi-step workflows, durable state, retries, approvals, and branching across services. It can surround an agentic process, but adding a state machine merely to inspect one Strands tool call would move the control outside the agent lifecycle and add orchestration the scenario says to avoid.
Question 19
Developers want one client abstraction for connecting an agent to any approved MCP server discovered from configuration. Which approach should the developer use? Choose ONE.
- Standardized function schema
- MCP client library
- Model-selection framework
- MCP for agent-tool interoperability
Correct Answer: B
Correct Answer
Answer B is correct because MCP client library — at this decision point, gives agents a consistent client-side, within this design, mechanism for discovering and invoking, within this design, MCP-exposed tools. Typical use: multiple MCP servers should be, in the described flow, accessed through a common client, in the described flow, pattern. Goal: standardize client-side discovery and invocation of MCP tools. MCP client library provides that behavior.
Incorrect Answers
Answer A is incorrect because A standardized function schema makes one tool’s callable contract clear to the model, but it is not a reusable client abstraction for discovering and connecting to different MCP servers. The requirement is common client-side protocol handling across approved servers, which is provided by an MCP client library.
Answer C is incorrect because Model-selection framework — for this design, applies policy or scoring logic, for this workload, to choose among candidate models, for this workload, at runtime. Supports designs where the application must choose a, for the stated need, model based on measurable request, for the stated need, characteristics. The workload needs standardize client-side discovery and invocation of MCP tools. Model-selection framework maps elsewhere.
Answer D is incorrect because MCP for agent-tool interoperability — within this implementation, uses the Model Context Protocol, for this decision, to expose tools through a, for this decision, consistent agent-to-tool interface. Strong fit when tools should be reachable through, within this design, a standardized protocol instead of, within this design, bespoke integrations. Operational need: standardize client-side discovery and invocation of MCP tools. MCP for agent-tool interoperability is secondary here.
Question 20
An agent loops through planning and tool use. The business rule says it must stop as soon as the order status becomes CONFIRMED. What should the team implement? Choose ONE.
- Lambda timeout guard
- Circuit breaker around failing tool
- Explicit stopping condition
- Idempotent tool operation
Correct Answer: C
Correct Answer
Answer C is correct because Explicit stopping condition — for the described workload, terminates iterative reasoning or tool-use, for the stated need, loops when a defined completion,, for the stated need, risk, or iteration condition is, for the stated need, reached. Best suited when an autonomous loop must not, for this decision, continue indefinitely. Deciding factor: terminate the loop on a defined business completion state. Explicit stopping condition fits this case.
Incorrect Answers
Answer A is incorrect because Lambda timeout guard — for the described workload, enforces bounded execution around a, for the stated need, tool or model-supporting function so, for the stated need, stalled work cannot run indefinitely. Intended when a lightweight serverless action needs, for this decision, a clear execution timeout. The architecture needs terminate the loop on a defined business completion state. Lambda timeout guard is a poor fit.
Answer B is incorrect because A circuit breaker opens because an external dependency is failing repeatedly; its trigger is service health, not business completion. The agent’s tools may all be healthy when the order becomes CONFIRMED. An explicit stopping condition should terminate the planning-and-tool loop as soon as that business state is reached.
Answer D is incorrect because Idempotent tool operation — in this scenario, uses an idempotency token or, in this case, state check so a retried, in this case, action does not create duplicate, in this case, side effects. Appropriate if workflow retries could otherwise repeat, at this layer, a charge, ticket, or update. The stem requires terminate the loop on a defined business completion state. Idempotent tool operation serves another purpose.
