Auto Ontology

Auto Ontology is NVIDIA’s open-source context layer for AI agents working with enterprise data, earning over 3,600 stars on GitHub. Released under the Apache 2.0 license, it runs as a service beside your databases and compiles their metadata into an ontology that agents can query reliably, instead of guessing from raw schemas.

Overview

Agents can retrieve documents and query databases, but reliability is what limits agentic analytics: a revenue figure delivered faster is worth nothing if it is wrong. Before an agent queries, it needs to know where the data lives, what it means, whether it should be used, and how it should be used. Auto Ontology answers those questions up front. It maps business terms to columns, attaches reviewed SQL for measures, and records the relationships between tables, so agents query through a shared, certified model of the data rather than improvising joins and measures for every question. A web interface lets people review, edit, and certify definitions, and the system complements semantic models teams already own.

Key Features

• Ontology discovery: generate an ontology from database metadata or import existing semantic models, then explore the result as a semantic graph.
• Grounding: give agents one definition per business concept, mapped consistently across tables and databases.
• Certification: review, edit, and certify definitions through the web UI or the API before agents rely on them.
• Answerability checks: validate whether a question can be answered before querying, so the agent asks for clarification instead of answering confidently and wrongly.
• Transparent answers: every answer returns the SQL used and the agent’s steps, making results auditable.
• Open agent access: serve context to any agent harness through MCP, agent skills, and a REST API.
• Measured quality: a companion evaluation repository benchmarks ontology-driven answers against schema-only prompts on datasets such as BIRD.

💳 Pricing

Auto Ontology is fully open source and free to self-host. Running it requires Docker and an NVIDIA API key for the underlying models, but the software itself carries no license fee.

Who Is It For

Auto Ontology suits data teams and developers building agentic analytics over enterprise databases, especially where wrong answers carry real cost. If your agents already query business data and you need their numbers to be explainable, certified, and reproducible, this is a strong foundation.