Kiro Cheat Sheet
-
Kiro is an agentic coding service from AWS that helps developers plan, write, test, and document software. It uses foundation models through Amazon Bedrock to interpret requests, work with project context, and carry out development tasks.
-
Use conversational coding to explore an unfamiliar codebase, explain a function, make a targeted change, or generate tests. Kiro can read relevant files, propose edits, and run commands when its permissions allow.
-
Use spec-driven development when a change needs more planning. Kiro helps turn a prompt into requirements with acceptance criteria, a technical design, and a task list you can review before implementation.
-
Kiro handles model inference and coordinates the agent’s work. You provide the project context, set access permissions, and review the resulting code and test results before merging.
Features
-
Conversational coding
-
Ask Kiro to explain a codebase, draft a small change, write tests, or update documentation from a natural-language prompt.
-
-
Spec-driven development
-
Define requirements and acceptance criteria, review the technical design, and break implementation into tasks. Kiro supports structured workflows for both features and bug fixes.
-
-
Development interfaces
-
Use the desktop IDE for editor-integrated work, the CLI for terminal and CI workflows, or Kiro Web for cloud sessions. The mobile interface lets you monitor tasks, review pull requests, and chat with the agent.
-
-
Repository and tool integrations
-
Kiro Web supports GitHub and GitLab repositories and can open pull or merge requests. Model Context Protocol (MCP) integrations let the agent use external tools and data sources.
-
Components
-
Specs organize work into requirements, design, and implementation tasks. A feature spec can start from requirements or from a technical design. A bugfix spec records the current behavior, the expected behavior, and what should continue working after the fix.
-
Steering gives the agent persistent project guidance, such as naming conventions, architectural decisions, and test commands. Store shared workspace guidance in
.kiro/steering/. -
Hooks run an agent prompt or shell command when a configured event occurs. For example, run a formatter after the agent changes a source file or update documentation after a spec task finishes.
-
MCP servers provide tools or data from external systems. Configure a server when the agent needs to query a service or use an integration.
-
Powers package task-specific guidance and can bundle MCP configuration. Kiro loads relevant guidance and tools when a task matches the power.
-
Cloud sessions run tasks in isolated sandboxes. Kiro clones authorized repositories into the session and retains the workspace so you can return to it later.
Kiro Security
-
Individuals can sign in with AWS Builder ID or supported social identities. Organizations can manage user access with AWS IAM Identity Center. AWS IAM policies govern administrator actions in the Kiro console. Kiro access and AWS resource permissions are managed separately.
-
Configure which files, shell commands, and MCP tools the agent can access. Review approval settings before granting persistent access to a tool or command.
-
Kiro documents TLS 1.2 or later for data in transit and AWS-owned AWS KMS keys for encryption at rest.
-
Kiro Web sandboxes run with configurable internet access. Limit reachable domains and authorize only the repositories a session needs.
-
Enterprise administrators can export per-user activity, configure prompt logging, and use AWS CloudTrail to audit supported API activity and agent tool calls. See the monitoring guide for event settings and coverage by client.













