Author here. I wrote this because I kept seeing teams solve agent context by stuffing prompts or calling N APIs per request—fine for demos, breaks at scale.
I wanted to put a name to the alternative I have in mind: a context plane—infrastructure that builds, stores, and serves context so agents get the right context per entity with one API (store + cache, entity-first, S3-compatible object storage, Rust).
I'd love feedback from the community: how are you solving context for agents today, and what would you recommend or do differently?
I wanted to put a name to the alternative I have in mind: a context plane—infrastructure that builds, stores, and serves context so agents get the right context per entity with one API (store + cache, entity-first, S3-compatible object storage, Rust).
I'd love feedback from the community: how are you solving context for agents today, and what would you recommend or do differently?