It comes with its own extension with auth, save images, videos and links from multiple sources. The saved items are stacked and analyzed by a vision model which tags them and goes through an embedding model to be saved in the database. You can connect your agent with mcp, inspect videos frame by frame and comment your ideas to share with your team.
Its insanely cheap to run ($0) and maintain, which is simply the cost of maintaining your library as it grows.
We have been using it for Hooped (
https://www.hooped.design/ ) for the past two months as we developed this and it has saved us hours of time during exploration when starting a project.