Neocarta is an end-to-end Python library for building a semantic layer in Neo4j. It ingests schema metadata, business glossaries, governance tags, and query history from your data sources into a unified graph — then serves that graph to AI agents through a Model Context Protocol (MCP) server, enabling reliable Text2SQL, query routing, and data discovery.Documentation Index
Fetch the complete documentation index at: https://mintlify.com/neo4j-labs/neocarta/llms.txt
Use this file to discover all available pages before exploring further.
Quickstart
Ingest a dataset and connect your first agent in under five minutes.
Connectors
Explore connectors for BigQuery, Dataplex, CSV, JDBC, Unity Catalog, Databricks, and OSI.
MCP Server
Expose your semantic layer to AI agents with catalog and search tools over stdio.
CLI Reference
Drive ingestion from the shell with the noun-verb Neocarta CLI.
How it works
Neocarta follows a simple three-step pattern: ingest, enrich, and serve.Ingest
Run a connector to extract schema metadata from your data source and load it into Neo4j. Only metadata crosses into the graph — your data stays in the source.
Enrich (optional)
Generate vector embeddings for table and column descriptions to unlock semantic search.
What the graph contains
The semantic layer unifies four categories of metadata in a single Neo4j graph:Schema Metadata
Tables, columns, data types, primary keys, foreign keys, and sample column values from your databases.
Business Glossary
Glossary terms and categories linked to the tables and columns they describe, sourced from Dataplex or CSV.
Governance Tags
Vendor-neutral governance tag definitions from Databricks or other platforms, associated with tables and columns.
Query History
Real SQL queries and the tables and columns they reference, revealing actual usage patterns across your data.
Supported data sources
BigQuery
Schema metadata and query logs from Google BigQuery.
Dataplex
BigQuery catalog and business glossary from GCP Dataplex Universal Catalog.
CSV Files
Load any structured metadata from CSV files.
JDBC
Any JDBC-compatible database via SchemaCrawler.
Unity Catalog
Open Unity Catalog REST API for schema metadata.
Databricks
Managed-Databricks Unity Catalog governance tags.
Neocarta is a Neo4j Labs project — experimental and community-supported. It requires a running Neo4j instance (AuraDB free tier, Neo4j Desktop, or Docker).