Based on repository activity, growth velocity and community engagement.
31
Growth4/30
Activity8/25
Popularity11/25
Trust9/20
368
Stars
High
Sentiment
Votes
368
README.md
Kiro Agent - VS Code Extension
Overview
The Kiro Agent is a sophisticated AI-integrated, spec-based development environment that transforms how developers build software. This VS Code extension enables developers to define systems and their components using friendly, natural language specifications that generate production-ready code for deployment.
Key Features:
AI-powered coding assistant with multi-provider support
Specification-based development workflow
Advanced prompt engineering system
Comprehensive tool ecosystem
Real-time collaboration capabilities
Enterprise-grade security and authentication
Architecture
Extension Structure
The extension follows a monorepo architecture with 11 specialized packages:
Template factory pattern for consistent message handling
Streaming Support:
Real-time response streaming via streamComplete() methods
AWS CodeWhisperer streaming client integration
Bidirectional communication with webview components
Prompt Engineering System
Base System Prompt
The core system prompt (getBasePrompt()) establishes Kiro's comprehensive identity and operational framework. This prompt defines Kiro's personality, capabilities, response style, and key features.
Kiro includes optimized edit prompts for 14+ different AI models and providers. Each prompt is carefully crafted to work with the specific formatting and instruction-following patterns of different models.
Kiro implements a sophisticated specification-based development workflow that guides users through requirements gathering, design, and implementation planning.
Multi-Modal Intent Detection:
The IntentClassifier uses sophisticated prompt engineering for intent recognition:
Classify the user's intent from their message:
- DO: Imperative actions (default mode)
- CHAT: Questions and explanations
- SPEC: Specification creation and management
Consider context, tone, and explicit keywords to determine intent.
Provide confidence scores for classification accuracy.
Classification Strategies:
Local Rule-Based: Pattern matching for common intents
LLM-Based: Sophisticated intent detection with confidence scoring
Hybrid Approach: Combines local and LLM results with weighted scoring
27 language parsers including JavaScript, TypeScript, Python, Rust, Go, Java, C++
Code snippet extraction queries
Import statement detection
Function/method definition parsing
Tokenizers:
Llama tokenizer with worker pool
Tiktoken for OpenAI models
Multi-threaded tokenization
Accurate token counting
Assets & Icons
Icon System:
495+ SVG icons for comprehensive UI coverage
Development tool icons
File type indicators
Custom Kiro branding
Development Workflow
Build System
Bundling Strategy:
ESBuild for main extension bundle
Selective external dependencies
Native binary preservation
Platform-specific optimizations
Build Commands:
npm run compile # Build all components
npm run package # Compile code only
npm run release # Full build with packaging
npm run analyze-externals # Update dependencies
Testing Framework
Test Configuration:
Mocha framework with XML reporting
Chai assertions and Sinon mocks
Test discovery in __tests__ directories
VS Code extension testing support
Extension Activation
Activation Events
File System Support:
Custom schemes: kiro-diff, kiro-meta, kiro-spec
Language detection: JavaScript, TypeScript variants
Webview activation: onboarding, GUI, hooks
Command Activation:
Startup commands and configuration
Execution triggers and event handlers
Language server integration
Key Bindings
Essential Shortcuts:
Cmd+L - Focus chat input
Cmd+I - Start inline chat
Shift+Cmd+Enter - Accept changes
Shift+Cmd+Backspace - Reject changes
Cmd+K Cmd+A - Toggle tab autocomplete
Advanced Features
VS Code API Integration
Experimental APIs:
21 experimental VS Code APIs
Chat participants and diff commands
Terminal integration and file search
Semantic tokens and inline actions
Custom Providers:
File system providers for custom schemes
Language servers for syntax highlighting
Webview providers for rich UI
Authentication providers for secure access
Performance Optimizations
Efficient Processing:
Streaming AI responses
Incremental parsing and analysis
Lazy loading of heavy components
Worker-based tokenization
Memory Management:
Context pruning for large conversations
Efficient bundling strategy
Resource cleanup and disposal
Optimized asset loading
Extensibility
Plugin Architecture
MCP Integration:
External tool discovery
Server lifecycle management
Protocol-level communication
Custom tool development
Hook System:
Event-driven automation
Custom trigger conditions
Flexible action definitions
Workspace-specific configuration
API Surface
Extension Points:
Custom tool definitions
Prompt template customization
Authentication provider integration
UI component extensions
Troubleshooting
Common Issues
Configuration Problems:
Check .kiro/settings/mcp.json for MCP server issues
Verify authentication tokens in ~/.aws/sso/cache/
Review extension logs for error details
Performance Issues:
Disable unnecessary MCP servers
Reduce context window size for large files
Enable development mode for detailed logging
Debug Tools
Logging System:
Extension output channel
LLM request/response logging
MCP server communication logs
Performance metrics tracking
Contributing
Development Setup
Clone the repository
Install dependencies: npm install
Build the extension: npm run compile
Launch VS Code with the extension: F5
Adding External Dependencies
Add package to UNBUNDLEABLE_PACKAGES in scripts/analyze-externals.mjs
This extension represents a comprehensive AI-powered development environment that combines the power of large language models with sophisticated tooling and workflow management to create a next-generation coding experience.
Ecosystem Role
Standard MoltPulse indexed agent.
Embed Badge
Show off your Pulse Score in your GitHub README to build trust and rank higher.
