jaturapornchai/mongomodeleditor

MongoModel is a browser based visual tool for designing MongoDB database structures without connecting to a real database. You lay out collections as draggable boxes on a canvas, draw relationship lines between them, describe each field, and then export the result as working code you can actually use, in formats like mongosh scripts, Mongoose schemas, or TypeScript interfaces. The interface uses a dark theme similar to DataGrip and is built around Thai language descriptions. Each collection box can be resized, renamed inline, duplicated, or deleted, and an auto layout button arranges related collections automatically. Fields support nine data types, including String, Number, Boolean, Date, ObjectId, Array, Object, Decimal128, and Mixed. You can mark a field as required, add a unique index, set an enum with a default value, and the tool requires every field and collection to have a description written in Thai, warning you if one is missing. Relationships are drawn by dragging from a field to another collection, and double clicking a connecting line cycles it between a reference or embedded relationship, with cardinality options of one to one, one to many, or many to many. Projects and their diagrams are saved automatically about every 400 milliseconds to a shared data file on the server, which acts as the single source of truth. Multiple diagrams can live inside one project as separate tabs, and changes made by another browser tab or by a connected AI assistant appear automatically within a few seconds along with a notification. The tool can export a diagram into seven formats: mongosh commands, Mongoose schemas, TypeScript interfaces, a Thai language data dictionary in Markdown, an Obsidian style linked wiki with a graph view, sample JSON documents, and the raw diagram data as JSON. It also runs a built in MCP server, so other AI tools such as Claude Desktop or Cursor can connect and read, add, or edit diagrams directly, though any description they write must also be in Thai. The project can be run with Docker, which is the recommended setup, or with Node.js version 20 or newer, and it has no authentication, since it is meant to run only on your own machine. The README notes that every line of code in the project was written by AI through a multi-agent process.