Kiro Agent is a VS Code extension that transforms software development into a spec-driven, AI-assisted workflow. It empowers developers to define systems using natural language, generating production-ready code with multi-provider AI support, advanced prompt engineering, and real-time collaboration.

How It Works

Kiro Agent employs a monorepo architecture with specialized packages for core AI chat, shared utilities, UI components, and integrations like AWS CodeWhisperer. It leverages a sophisticated prompt engineering system, supporting numerous LLMs (OpenAI, Anthropic, Mistral, Gemini, Ollama, Amazon Q) with model-specific prompt templates and dynamic context injection. The system uses a hierarchical prompt structure, combining base system prompts, model formatting, context, and tool-specific instructions for robust AI interaction.

Quick Start & Requirements

• Install: Clone the repository and run npm install followed by npm run compile. Launch VS Code with the extension using F5.
• Prerequisites: Node.js, npm.
• Links: GitHub Issues

Highlighted Details

• Supports 14+ AI models with specialized edit prompts and a template factory for consistent message handling.
• Features a specification-based development workflow with prompts for requirements clarification, design, and implementation planning.
• Integrates a multi-modal intent detection system using local rules and LLM-based classification.
• Includes built-in tools for file system operations, search, and shell command execution, with MCP integration for external tools.
• Utilizes Tree-sitter for parsing 27 programming languages and supports various tokenizers for accurate token counting.

Maintenance & Community

• Project managed by ghuntley/amazon-kiro.kiro-agent-source-code-analysis.
• Support and issues can be tracked via GitHub Issues.

Licensing & Compatibility

• Licensed under the Apache-2.0 License.
• Permissive license suitable for commercial use and integration with closed-source projects.

Limitations & Caveats

The extension relies heavily on VS Code's experimental APIs, which may lead to instability or breaking changes. While it supports local model execution via Ollama, performance will be dependent on user hardware.