Coolver/home-assistant-vibecode-agent

What it solves

It eliminates the need for manual YAML editing and deep documentation searches when managing Home Assistant. It allows users to describe smart home goals in natural language, which an AI then translates into functional automations, scripts, and dashboards without the user having to manually write the code or copy-paste snippets.

How it works

The project uses a two-part architecture to provide a stable interface for AI IDEs:

1. Home Assistant Agent: A service running inside or alongside Home Assistant (as an Add-on or Docker container) that has native access to APIs, files, and runtime state.
2. Home Assistant MCP Server: A module running on the user's computer that uses the Model Context Protocol (MCP) to connect the AI IDE (like Cursor, Claude Code, or VS Code) to the Agent via a secure REST API.

This design avoids fragile SSH-based scripting by providing the AI with structured, relevant context and a controlled set of actions to modify the configuration safely.

Who it’s for

Experienced Home Assistant users who want to accelerate their automation workflow, as well as those who prefer using AI to design and deploy complex smart home systems while maintaining the ability to review and roll back changes.

Highlights

• Natural Language Deployment: Create complex systems (e.g., climate control with multiple TRVs) by simply describing the requirement.
• IDE Integration: Works directly with MCP-enabled IDEs like Cursor, VS Code, and Claude Code.
• Safe Modifications: Includes automatic Git-based versioning for all changes, allowing for easy rollbacks.
• Comprehensive Management: Can analyze entities, create helpers, design UI dashboards, tweak themes, and manage HACS integrations.
• Native Access: Uses a dedicated agent to access internal HA APIs and state rather than relying on external SSH hacks.