Kiro — Technical Documentation Highlights

Spec-Driven Development Workflow

Kiro's core workflow transforms a natural-language prompt through three automated stages:

Requirements — structured, verifiable statements of what the system must do.
Architecture — a design plan derived from those requirements.
Tasks — discrete, executable implementation units assigned to agents.

Automated reasoning runs over the requirements stage to surface contradictions and coverage gaps before any implementation begins.

Agentic Execution

Local agents run within the developer's environment.
Cloud agents execute in parallel, enabling concurrent workstreams.
Hooks allow developers to define event-driven background automations that trigger without manual intervention.

CLI & CI/CD

Kiro ships a headless CLI that exposes agentic capabilities outside the IDE, enabling integration into any CI/CD pipeline. Shell support covers bash, zsh, and fish.

Protocol Compatibility

MCP (Model Context Protocol) — connect external context sources and tools.
ACP (Agent Client Protocol) — interoperate with external agent runtimes.

IDE & Extension Ecosystem

Kiro is compatible with VS Code settings and the Open VSX extension registry, giving access to a broad ecosystem of editor extensions without migration friction.

Integrations Reference

Integration	Category
Figma	Design handoff
Terraform	Infrastructure as code
MCP	Context & tooling protocol
ACP	Agent interoperability
Open VSX extensions	Editor ecosystem
VS Code settings	Editor configuration
bash / zsh / fish	Shell environments
CI/CD	Automated pipelines

Model Configuration

Kiro supports an Auto routing mode alongside explicit selection of Claude (Anthropic), GPT (OpenAI), DeepSeek, MiniMax, GLM, and Qwen (open-weight). Model selection can be configured per task or globally.

API

Kiro exposes a public API for programmatic control of agentic workflows and automation.

Security

The platform is built on AWS infrastructure, providing enterprise-grade security controls suitable for regulated and large-scale engineering environments.