Key Concepts
- Kirao: Amazon's new AI IDE in preview phase.
- Aentic AI IDE: AI-powered Integrated Development Environment.
- VS Code fork: Kirao is built upon the VS Code platform.
- Agent Hooks: Automate routine tasks based on event triggers.
- MCP Integration: Direct connections to APIs, databases, and internal tools.
- Spec Mode: Structured, in-depth planning and execution for tasks.
- Vibe Mode: Autonomous generation for quick prototyping and chat-based tasks.
- Context Engineering: Providing detailed context and instructions for AI models to reduce hallucination.
- Inline Edits: Making edits directly within the chat or editor.
1. Introduction of Kirao: Amazon's AI IDE
- Amazon has introduced Kirao, a new Aentic AI IDE, joining the AI IDE landscape.
- Kirao is currently free during its preview phase.
- It is a VS Code fork with a polished interface and features like hooks and code with spec mode.
- Kirao aims to assist in prototyping apps to production.
2. Key Features and Functionalities
- Spec-Driven Deployment: Kirao transforms prompts into requirements, designs, and linked tasks, mitigating AI hallucination by providing structured guidelines.
- Agent Hooks: Automate routine tasks (e.g., generating tests, writing documentation, optimizing code) triggered by events like file saves.
- Multimodal Interface: Supports uploading UI screenshots or architecture diagrams for Kirao to interpret and guide coding decisions.
- MCP Integration: Enables direct connections to APIs, databases, and internal tools for full-stack awareness.
- Modes of Operation:
- Spec Mode: Requires more planning, using in-depth analysis for structured and efficient task execution; suitable for prototyping for deployment.
- Vibe Mode: Designed for quick generation, chat-based tasks, Q&As, and explanations.
3. Installation and Setup
- Kirao can be downloaded from the provided link for Mac OS, Windows, or Linux.
- Users can sign up/sign in using Google, GitHub, AWS builder ID, or organization identity.
- Import configurations from VS Code (optional) to install necessary extensions and themes.
- Shell integration is crucial for IDE functionality.
4. Using the Kirao Interface
- The primary interface is the chat tab, similar to other AI IDEs.
- The base command set provides access to the chat tab.
- Users can add context by including images and using the Cloud Sonnet 4.0 model.
- Autopilot mode can be toggled for human-in-the-loop generation.
- A dedicated Kirao tab manages specs, agent hooks, agent steering, and MCP servers.
5. Agent Hooks, Steering, and MCPs
- Specs: Define structured ways for the AI to build applications by adding descriptions.
- Agent Hooks: Allow the agent to listen to file changes and trigger actions, ensuring adherence to features and reducing hallucination.
- Agent Steering: Acts as a rulebook to ensure the AI agent is steered in the right direction.
- MCPs: Managed via a JSON file to configure development processes easily.
6. Demonstrations and Examples
- CRM Dashboard Generation: Demonstrates both Vibe and Spec modes.
- In Vibe mode, the AI autonomously generates a CRM dashboard.
- Checkpoints allow users to revert to previous states.
- Spec Mode for CRM Dashboard Improvements:
- The AI reads files and creates a requirement sheet, ensuring structured improvements.
- Users can edit requirements in real time.
- The process includes requirement, design, and implementation phases.
7. Inline Edits and Commands
- Inline edits can be performed by highlighting code and using
ctrl + lto add a reference to the panel, orctrl + ito make direct edits within the editor. - A list of commands can be accessed via
ctrl + shift + p, offering options for debugging and code generation.
8. The Design and Implementation Phases in Spec Mode
- After approving the requirements, Kirao generates a detailed design plan.
- Users can refine the plan.
- The final step is the implementation plan, which involves coding the application using the Sonnet model.
- Generating a detailed plan with the Spec mode leads to higher-quality outputs.
9. Synthesis/Conclusion
Kirao is presented as a promising AI IDE by Amazon, offering unique features like spec-driven development, agent hooks, and MCP integration. It caters to both rapid prototyping (Vibe mode) and structured development (Spec mode). The emphasis on context engineering, detailed planning, and user-controlled checkpoints aims to mitigate AI hallucination and improve the quality of generated code. The IDE's free preview offers an opportunity for users to explore its capabilities and provide feedback.
AI summaries can miss context or contain errors. Check important details against the original video.





