DOCSEARCHbyAlgolia
Submit Docs

Kiro Documentation

website·Indexed 41 days ago

https://kiro.dev

Documentation for Kiro, featuring guides on installation, custom agents, and the Model Context Protocol (MCP). It covers the use of specs (Feature, Bugfix, and Quick Specs), steering, hooks, and capability-based permissions for managing agent access. Includes details on available models, reasoning effort settings, and workflows for requirements-first and design-first development.

Tokens
199.6K
Snippets
795
Records
1.6K
Agent score
74%

What's inside Kiro

  1. Overview of Kiro Code Intelligence capabilities

    Kiro provides two layers of code understanding. Tree-sitter is built-in and works across IDE, CLI, Web, and Mobile without installation. LSP integration is optional, requires a language server installation, and is available only on IDE and CLI for higher precision tasks like refactoring and diagnostics.

  2. Overview of the Kiro Bugfix Spec workflow

    The Bugfix Spec is a structured workflow designed to diagnose and fix bugs with surgical precision. It prevents regressions by explicitly documenting what should change and what must remain unchanged. The workflow consists of three phases:

    1. Bugfix Analysis Phase (bugfix.md): Defines the current defect, the expected correct behavior, and the unchanged behavior that must be preserved to prevent regressions.
    2. Design Phase (design.md): Kiro performs root cause analysis and proposes a fix approach, defining properties to test for the bug's existence, the fix's success, and the preservation of existing functionality.
    3. Tasks Phase: Implementation tasks are generated using property-based tests (PBTs) to validate the fix and ensure no regressions were introduced.

  3. Overview of Crew Knowledge Library

    The Knowledge Library is a curated store of external content (files, folders, and URLs) that you intentionally provide to the agent for reference. Unlike Memory, which is learned automatically from conversations, Knowledge is manually managed. It is used for project documentation, API references, runbooks, SOPs, and code repositories. Crew chunks, indexes, and embeds this content so the agent can retrieve relevant snippets during conversations.

  4. Overview of Kiro built-in agent tools

    Kiro's agent provides a set of built-in tools for development tasks across different surfaces (IDE, CLI, Web, and Mobile). While the default agent has access to all tools, custom agents can restrict access via configuration.

  5. Overview of Kiro Web

    Kiro Web is a browser-based AI development agent accessible via app.kiro.dev. It allows developers to build software through conversation, write code, and open pull requests or merge requests. Kiro never merges changes automatically; users must review results before they reach the repository's default branch.

  6. Overview of Kiro Cloud Sessions

    Cloud sessions run the Kiro agent harness in a managed cloud sandbox instead of on the local machine. These sessions are tied to the Kiro account, allowing users to start a task on one device (e.g., Web) and resume it on another (e.g., IDE or CLI). The agent continues working in the sandbox even when the user is disconnected. Cloud sessions are included in existing Kiro plans with a limit of 10 concurrent sessions.

  7. Overview of Artifact Deploy

    Artifact Deploy allows users to deploy kind: "webapp" artifacts to their own AWS account, providing a public HTTPS URL. Crew orchestrates the deployment, but the user's AWS account is billed for the actual resources used. The system supports three deployment tiers based on the artifact's needs: Static (S3 + CloudFront), Fullstack (adds Lambda Function URL), and Stateful (adds DynamoDB).

  8. Overview of Kiro per-user activity monitoring

    Kiro allows administrators to export daily per-user activity telemetry to audit usage, optimize license allocation, track credit/overage consumption, and analyze user interaction across clients. There are two independent export formats available: CSV files delivered to Amazon S3 (ideal for queries and dashboards) and OpenTelemetry (which pushes aggregate metrics to an OTLP-compatible collector). Both formats provide the same daily aggregate usage metrics.

  9. Overview of Kiro Permission Policies

    Permission policies allow administrators to deploy managed rules that restrict what the Kiro agent can do across an organization. These rules take precedence over a user's personal settings.

    Conflict Resolution: When an admin rule and a user rule target the same capability, the most restrictive effect wins: deny > ask > allow.

    Constraints:

    • Admin policies can only use deny or ask; they cannot use allow to grant permissions.
    • Policies are enforced at the client level and can be circumvented by users with administrative access to their local machine.

  10. Overview of Kiro Feature Specs

    Feature Specs provide a structured framework for building new features by guiding developers through requirements gathering, technical design, and implementation planning. They are best suited for complex features with multiple tasks, projects requiring team collaboration documentation, or features needing iterative design. They are not intended for bug fixes (which should use Bugfix Specs) or exploratory coding.

  11. Overview of Kiro built-in agents

    Kiro provides several pre-configured agents tailored for specific development workflows. These agents can be switched during a session or invoked via UI controls and slash commands.

  12. Overview of Model Context Protocol (MCP) in Kiro

    The Model Context Protocol (MCP) allows Kiro to connect to specialized external servers to access domain-specific tools, prompts, and resources. This enables Kiro to integrate with external APIs, access specialized knowledge bases (e.g., AWS Documentation), and use server-provided prompt/resource templates via the # mention system in chat.