Catalogs manage live metadata: assets, lineage, owners, glossary, and policies.
Why rawctx
Why rawctx?
Don't make agents look up meaning live. Release the meaning they should use.
The adjacent layers
Most tools around AI data systems fall into three layers.
Semantic runtimes execute definitions: they generate SQL, serve metrics, and query warehouses.
LLMOps tools trace prompts, model calls, retrieval, and evals.
rawctx layer
rawctx is a different layer.
rawctx releases business meaning as a versioned semantic package. A package can be reviewed, diffed, downloaded, loaded in Python, and rendered into LLM-ready context without querying your warehouse.
The goal is not to replace your catalog or semantic runtime.
The goal is to control the meaning variable.
Release the exact business meaning agents should use instead of making them infer it live.