We’ve been working on Habits, an open-source system for building automations, AI agents, internal tools, and full apps from the same graph-based model.
The core idea is that you design logic as nodes we call Bits, optionally attach a UI, export the result as YAML, and then package it for different targets depending on the use case. That means the same project can become a Docker deployment, a standalone binary or single executable, a desktop app, or a mobile app.
We built it because we kept running into the same gap. Workflow tools are great for automation, but they usually stop short of helping you ship complete end-user apps. App builders help with UI, but often tie you to a specific platform or runtime. AI-generated code is fast, but not always something we wanted to maintain or hand over to teams. We wanted a system where the same project could move between code, visual editing, and AI-assisted generation while still being exportable and self-hostable.
The project has two main parts. Base is the visual builder for logic and UI, while Cortex is the runtime that executes the exported app or workflow. A Habit can be created by writing YAML files directly, building visually in Base, or generating it with AI.
I’d especially love feedback on the architecture, the export and packaging model, and whether this feels more useful as an automation tool, an app builder, or something in between.
We’ve been working on Habits, an open-source system for building automations, AI agents, internal tools, and full apps from the same graph-based model.
The core idea is that you design logic as nodes we call Bits, optionally attach a UI, export the result as YAML, and then package it for different targets depending on the use case. That means the same project can become a Docker deployment, a standalone binary or single executable, a desktop app, or a mobile app.
We built it because we kept running into the same gap. Workflow tools are great for automation, but they usually stop short of helping you ship complete end-user apps. App builders help with UI, but often tie you to a specific platform or runtime. AI-generated code is fast, but not always something we wanted to maintain or hand over to teams. We wanted a system where the same project could move between code, visual editing, and AI-assisted generation while still being exportable and self-hostable.
The project has two main parts. Base is the visual builder for logic and UI, while Cortex is the runtime that executes the exported app or workflow. A Habit can be created by writing YAML files directly, building visually in Base, or generating it with AI.
The project is released under Apache 2.0.
GitHub: https://github.com/codenteam/habits
Docs: https://codenteam.com/intersect/habits
Video: https://www.youtube.com/watch?v=uhim-Y7b1vA
I’d especially love feedback on the architecture, the export and packaging model, and whether this feels more useful as an automation tool, an app builder, or something in between.