Amazon Bedrock Agents were designed to help generative AI applications move beyond answering questions and complete multi-step tasks. An agent interprets a user request, decides whether it needs more information, queries a knowledge base, or invokes an action such as calling an application API.
What makes an agent different from a chatbot?
A basic chatbot may generate a response from a prompt. An agent can orchestrate a sequence of steps: interpret the request, collect missing parameters, retrieve relevant information, call an API, inspect the result, and continue until it can respond or needs more input. The foundation model helps decide what step to take, while configured tools and permissions determine what the agent can actually do.
Core components of a Bedrock Agent
- Foundation model: interprets user input and supports orchestration and response generation.
- Instructions: describe the agent’s role, goals, boundaries, and expected interaction with users.
- Action groups: define operations the agent can perform. An action group can use an API schema such as OpenAPI and a Lambda function to fulfill the requested operation.
- Knowledge base: supplies relevant information that the agent can retrieve to improve responses.
- IAM service role: grants the agent the permissions required to invoke the configured resources.
- Guardrails: can be associated to filter or constrain inputs and outputs according to configured policies.
How runtime orchestration works
- The application invokes the agent with the user’s request.
- The agent interprets the request and determines what information or action is needed.
- If required parameters are missing, the agent can ask the user for them.
- The agent can query an associated knowledge base or select an action in an action group.
- The action is fulfilled through the configured integration, commonly a Lambda function calling an application API.
- The agent evaluates the returned observation and may continue the orchestration loop before producing a final response.
Action groups vs. knowledge bases
| Component | Purpose | Example |
|---|---|---|
| Action group | Perform an operation in another system | Check an order, create a ticket, or submit a booking request through an API |
| Knowledge base | Retrieve information to ground a response | Find a policy paragraph or product specification |
Build time and runtime
At build time, configure the model, instructions, action groups, knowledge bases, permissions, and any guardrail settings. Prepare the agent so it can be tested. At runtime, invoke it through the agent runtime API. Tracing can help troubleshoot the sequence of steps, tool invocations, knowledge-base queries, and observations.
Exam traps
- An action group is for taking an action; a knowledge base is for retrieving supporting information.
- The agent does not automatically have permission to call every API. IAM and the configured integration must authorize the operation.
- If the agent lacks required parameters, it may ask the user rather than guess.
- Agent orchestration is iterative; it may call tools and inspect their outputs before returning a final answer.
- Remember the current availability notice for Agents Classic and the AWS direction toward AgentCore.
Test yourself
Question 1
A Bedrock agent must create a support ticket by calling a company’s ticketing API. Which component defines the operation and its required parameters?
- Knowledge base
- Action group
- Model card
- Vector embedding
Reveal answer
Correct answer: B. Action groups define the operations an agent can perform and the parameters needed to fulfill them.
Question 2
An agent needs to look up a company’s internal return policy before answering a customer. Which component is designed to retrieve that information?
- Action group
- Knowledge base
- IAM access key
- Model Registry
Reveal answer
Correct answer: B. A knowledge base retrieves relevant information; an action group invokes an operation.
Key takeaway
Think of the classic Bedrock agent as an orchestrator: the model interprets, instructions guide behavior, knowledge bases retrieve context, action groups invoke operations, Lambda or another configured integration fulfills actions, and IAM controls access. Check AWS’s current AgentCore guidance for new architectures.
Related AWS topics
Amazon Bedrock Knowledge Bases · Amazon Bedrock Guardrails · Foundation Model Hallucinations
✦ CloudExamPro Premium
Ready for the AWS Certified AI Practitioner (AIF-C01)?
Move beyond memorizing terms. Practice with realistic, scenario-based questions and detailed explanations that help you understand the AWS AI services and concepts behind each answer.
Original practice questions. No dumps. No shortcuts.