Data Comprehension

Coming soon

A local comprehension engine for your files — point it at a folder and it reads everything in place, building a structured understanding of what your collection is and what each file means.

Data & Knowledge

Spaces — each context space is a directory read in place and comprehended into a structured understanding of the whole collection

Explore — browse every comprehended file and what the engine understood it to mean, with embeddings for retrieval

Findings — quality incidents across coverage, confidence, freshness, distribution and lineage, each traceable back to its evidence

Point it at any directory and it reads your files in place — never copying or moving them — comprehending each at ingest into a structured "context space" backed by local SQLite
A tiered pipeline builds understanding top-down — what the whole collection is, then what each file means — with embeddings for retrieval and optional web research to fill the gaps
Quality incidents surface what's weak — coverage, confidence, freshness, distribution and lineage — each with provenance back to the evidence, so you can trust or challenge what it concluded
Local-first — an MCP tool surface exposes the whole comprehension to your agents with a propose→review→commit write flow; it runs on your own machine against your own model via model-dock, no cloud