velum-labs/atlas — reverse-engineered prompt

Build me a Python tool called Alma Atlas that helps people discover what is inside their company data stack. I want a simple command line app that can connect to sources like BigQuery, Postgres, Snowflake, dbt, Airflow, Looker, Fivetran, and Metabase, scan them, and store the results in a local SQLite graph.

It should let someone search for tables, inspect schemas, see lineage, and understand what dashboards or downstream assets depend on a dataset. It also needs an MCP server so Cursor or Claude can ask questions like “what depends on this orders table” and get real context instead of guessing.

Please include a bundled sample stack so someone can try it without real credentials, with commands to install the sample, preview it, scan sources, and register the MCP server in Cursor or Claude. Add Docker support, docs, and safe handling for local credentials. Look up current MCP docs online if you need to.