Brood is a promptless, reference-first AI visual canvas for developers.

  Instead of wiring node graphs or writing long prompts, you import 1–3 reference images and run visual “abilities” directly on canvas. Brood keeps the workflow image-first and model/provider-flexible.

What it does now:

  - Single-image actions: diagnose, recast, variations, background edits, crop
  - Two-image actions: combine, swap DNA, bridge, argue
  - Ambient intent discovery: background intent classification while editing, with subtle visual nudges
  - Reproducible runs: artifacts + events written to disk for traceability

Why I built it: I wanted a faster “think with images” loop for product/creative workflows without heavy graph setup, while still keeping reproducibility and provider routing first-class.

Tech:

  - macOS desktop app (Tauri)
  - Python engine + CLI
  - Multi-provider model routing
  - Open source (Apache-2.0)

Repo: https://github.com/kevinshowkat/brood

I’d love feedback on:

  - where this beats/loses to node-based tools
  - highest-value workflows to prioritize next
  - what would make this a daily tool for you