AI-assisted development with Kiro: DVA-C03 study notes and practice questions
What to know about AI-assisted development with Kiro for the DVA-C03 exam, from the Development with AWS Services domain (30% of the exam guide), then 2 free questions to test it. Part of the DVA-C03 study guide.
Questions here ask which AI development tool fits a team’s workflow, and how much control the developer keeps. Kiro is the in-scope IDE for this task, so learn its features well enough to match each to a scenario.
Changes to the tooling
Amazon Q Developer IDE plugins closed to new customers on May 15, 2026, and end of support is April 30, 2027. Kiro replaces them.
Amazon Q Developer in the AWS console is unaffected. Older practice questions that name the IDE plugin describe the retired product, so match the scenario to Kiro’s features instead.
Kiro features
Specs. Kiro turns a prompt into requirements, a design and a task list before it writes code. Use specs for larger features where the developer wants to review the plan first. The spec files are requirements.md (user stories with acceptance criteria in EARS notation), design.md and tasks.md, and tasks can be run and tracked one at a time. This is the “spec-driven code generation” in the exam guide.
Steering files. Markdown files in the project that give the agent standing context: product purpose, technology choices, coding standards and code layout. Workspace steering files live in .kiro/steering/ and travel with the repository, so the whole team shares them; global steering files apply only on one developer’s machine. Use them when the agent must follow team conventions every time.
Agent hooks. Actions that run automatically on an event, such as a prompt being submitted, the agent finishing a turn, a tool call, or a file being created, saved or deleted. A hook can, for example, update tests or check for secrets when code changes. File Create, File Save and File Delete triggers fire only on changes the agent makes, not on a developer’s manual edits in the editor. A Pre Tool Use hook runs before the agent calls a tool, such as a file write, and a shell command that exits with a non-zero code blocks the call.
MCP servers. Connect the agent to external tools and data, such as a documentation server or a ticket system, using the Model Context Protocol. Servers are defined under mcpServers in .kiro/settings/mcp.json: the workspace file in the repository gives every developer the same tools, while the user-level file stays on one laptop.
Checkpoints. Kiro records a checkpoint for each prompt. Restoring one undoes the agent’s later file changes and the later chat.
Supervised and autopilot modes. Autopilot is the default: the agent creates and edits files and runs commands without stopping for approval, and the developer can view all changes as a diff or revert them afterward. In supervised mode Kiro stops after each turn that edits files, and the developer accepts or rejects each change, down to individual hunks. Choose supervised when every change needs human approval.
How to choose
Scenario wording
Answer
Plan and review a feature before code is written
Spec
Every generated change must follow team standards
Steering files
Run a check or update automatically after the agent saves a file
File Save agent hook
Connect the agent to an internal tool or data source
MCP server
Developer must accept or reject each change
Supervised mode
Block the agent from writing a secret before the file changes
Pre Tool Use hook
Undo the last few prompts’ changes and conversation
Restore a checkpoint
Upgrade a large code base with minimal review during the run
Autopilot, reviewed afterwards
Review generated code for secrets or style issues on every agent change
Agent hook that runs the review
Common traps
Recommending the Q Developer IDE plugin for a new project. Use Kiro.
Confusing steering with hooks. Steering supplies context; hooks run actions on events.
Expecting a File Save hook to run on manual edits. File hooks respond only to the agent’s changes.
Choosing autopilot, the default, when the scenario requires human approval of each change.
Assuming AI tools replace testing. The output still needs tests and review in the pipeline.
A developer has joined a payments team and must add a refund feature to an existing service.
The team wants the feature planned before anyone writes code. The developer needs an agentic AI IDE that turns the feature idea into requirements, a design and a task list, and then helps write the code.
The team also wants automations that run when the AI agent changes files, such as reminders to update documentation.
Which tool meets these requirements?
Choose one answer.
AA. Amazon Bedrock, a managed service that gives application code API access to foundation models for text and chat responses.
BB. Kiro, an agentic AI IDE that turns a feature idea into requirements, a design and tasks before any code is written, and runs hooks when its agent changes files.
CC. AWS CodePipeline, a continuous delivery service that builds, tests and deploys code through stages after the code is pushed to a repository.
DD. AWS Step Functions, a workflow service that coordinates AWS services through steps defined in a state machine.
Show the answer and explanation
Answer: B
A. Wrong: Bedrock gives application code API access to models. It does not run inside the IDE, plan a feature as specs, or run automations when the agent changes files.
B. Right: Correct. Kiro's specs turn a feature idea into requirements, a design and tasks before implementation starts. Its hooks run automations when the agent changes files, which covers the documentation reminders. Learn more: Kiro documentation (opens in a new tab)
C. Wrong: CodePipeline automates release stages after code is pushed. It does not plan a feature, write code inside the editor or react when an AI agent changes files.
D. Wrong: Step Functions runs workflows that call AWS services. It does not plan a feature with requirements and tasks, and it does not work inside the developer's IDE.
Explanation
Kiro is an AI-powered development environment from AWS. Its specs produce requirements, a design and tasks before code is written, and its hooks run automations when the agent changes files. Bedrock, CodePipeline and Step Functions each do a different job, and none of them plans a feature inside the IDE.
A platform team builds its services in Kiro, and every developer opens the same Git repository in Kiro. Kiro gives outdated answers about AWS service quotas, because it cannot read the current AWS documentation. The team also wants Kiro to read tickets from its internal ticket tool.
The team wants every developer who opens the project to get both tools automatically, without editing a file on their own laptop. The setup must be stored with the code.
Which change meets these requirements?
Choose one answer.
AA. Add the MCP server entries to the ~/.kiro/settings/mcp.json file on each developer's laptop, and ask each developer to copy the file.
BB. Paste the AWS documentation pages and the ticket list into the Kiro chat at the start of each session.
CC. Add the MCP server definitions under mcpServers in .kiro/settings/mcp.json at the project root, and commit that file to the repository.
DD. Create a Kiro agent hook on the File Save trigger that sends the AWS documentation to the agent each time a file is saved.
Show the answer and explanation
Answer: C
A. Wrong: This works on each laptop, but it is manual setup that the repository does not carry. The requirement is that developers get the tools automatically when they open the project.
B. Wrong: Pasted text is copied by hand, goes out of date and is limited to what fits in one chat. It does not give Kiro live access to the documentation or the tickets.
C. Right: Correct. Kiro reads workspace MCP settings from .kiro/settings/mcp.json, so every developer who opens the repository gets the same servers. Workspace settings take precedence over the user file when both define a server. Learn more: Configuring MCP servers in Kiro (opens in a new tab)
D. Wrong: File Save hooks run when the Kiro agent saves or modifies matching files, not when a developer needs information. Pushing text into the conversation on each save does not connect Kiro to live documentation or tickets.
Explanation
Kiro reads MCP server definitions from mcpServers in .kiro/settings/mcp.json. A workspace file in the repository gives every developer who opens the project the same tools, with no manual setup. The user-level file lives on each laptop, so it cannot be relied on for a shared setup.
Want more on AI-assisted development with Kiro? The 5 full-length practice sets have 6 more AI-assisted development with Kiro questions, 390 in all, none of them in the free exam. One payment of $14.99, no subscription. See the practice sets →