Marquee drafts a month of nights for an independent venue. Give it the neighborhood and a few things your regulars love, and it asks a taste graph which artists, films, shows, games, books and brands share that taste in that place.
- Resolve. It looks up the neighborhood, the venue and each anchor by name to get stable IDs.
- Ask. For each kind of night you picked, it requests recommendations that use the anchors, the location and any optional audience signals together.
- Adjust. If a list comes back thin, it relaxes one constraint at a time and tells you which.
- Schedule. It lays the best picks onto real dates, pairs each with a partner brand that shares its tags, and keeps two or three alternates ready.
Every pick shows why it was chosen. Swapping, locking and regenerating happen in your browser from the same candidate lists, so they are instant.
The agent is deterministic: no language model is needed to plan. Taste data comes from the Qloo API when this server has a key; otherwise it runs on a small invented dataset that is clearly labelled.
Built by AI coding agents directed by Keven Crick, with independent AI review and automated tests. Open source under the MIT license.