ochakai
A context provider for data agents that keeps metric definitions, verified example queries and interpretation notes in one knowledge base, served over MCP, REST and CLI.
ochakai fills the gap left by a semantic layer, which can say revenue equals a sum of prices but not whether a given number is good or bad, by keeping metric definitions, verified golden queries, interpretation knowledge on how to read a metric, glossary terms and table-catalog entries in one knowledge base, curated by humans and agents together and served to Claude Code and other data agents over MCP, REST, a CLI and a bundled web UI.
It runs as one Go binary plus Postgres, using Docker locally and Cloud Run with Cloud SQL in production. A public demo needs no account and accepts plain search and lookup requests over HTTP; text drawn from the warehouse, such as column and status names and SQL snippets, stays in its original language inside an otherwise localized base, so a search still reaches the right entries.
The public demo is a restorable sandbox that resets on a schedule and accepts writes, so users can try the intended workflow of an agent drafting a knowledge entry and a human approving it.