Product overview
Built for how Snowflake teams actually work
One real-time context layer, published everywhere your team and your agents look.
Architecture
Snowflake → AZ Data → Git + MCP
Three small components. The warehouse is only ever read by the poller, on a tight schedule.
DDL poller
Hash-diffs INFORMATION_SCHEMA every 1–5 minutes with a least-privilege role. No ACCOUNT_USAGE lag, no full scans.
Schema trackingRenderer
Turns every change into Markdown, OKF, and a graph, then commits to GitHub or Bitbucket with a readable message — or writes the same files to an S3 bucket or a Confluence space.
Git publishingMCP server
Serves structured, annotated context to Claude, Cursor, and your own agents — with a last-synced timestamp on every answer.
Agents & MCPSee it on your own schema.
Live walkthrough of the poller, the Git commits, and an MCP session — against your account if you'd like.