Compatibility
Minecraft: Java Edition
Platforms
Supported environments
Links
Tags
Creators
Details
Brick by Brick Tab
Brick by Brick Tab is a client-side Fabric mod that suggests the next block while you build. It is not an auto builder. You still place every block yourself, the mod only shows possible next steps.
The suggestion appears as a ghost block under the crosshair. The HUD can also show a few alternatives if the first one does not fit what you had in mind.
How it works
The mod uses a trained next-block model. It reads the recent blocks placed by the player, then predicts where the next block is likely to go, which block it should be, and how it should face.
Only player-placed blocks are used as context. Terrain, ores, plants, fluids, and blocks generated by the world are ignored.
The current model is downloaded from the project server and can update automatically when a newer version is published. The model is optional. If weights are unavailable, Brick by Brick Tab still provides basic line-continuation suggestions.
Features
- Ghost block preview under the crosshair
- Ranked suggestions on the HUD
- Green preview when the suggested block is available in the inventory
- Red preview when it is not available
- Local adaptation to the way you build
- Client-side only, no server installation required
- No automatic placement or block spawning
Controls
- B toggles suggestions
- V switches to the next suggestion
- Other actions can be assigned in the Controls menu
Train your own model
You can train and use your own model. The repository contains the model specification, weight format, palette format, and training notes. A custom model needs a model.bbbt file and a palette.json file.
See the source code and training documentation on GitHub.
Planned
The trained model will continue to receive new versions and better training data. One of the planned directions is text to build, where a build can be started from a text description. This is planned work and is not available yet.
Data and privacy
Placed-block data stays on the device by default. Sharing placement data for future training and sharing build descriptions are separate opt-in settings. Nothing is uploaded without consent.


