// the find
open-metadata/OpenMetadata
The Open Context Layer for Data and AI , OpenMetadata is the open platform for building trusted data context and business semantics for humans, AI assistants, and agents.
OpenMetadata is an open-source data catalog and metadata management platform that connects technical metadata, lineage, data quality, governance, and business semantics into a unified graph. It's aimed at data engineering teams who need a central place to track what data exists, who owns it, whether it's trustworthy, and how it flows through pipelines. Recent pivot toward positioning itself as an 'AI context layer' with an MCP server adds a legitimate use case for grounding LLM agents in governed enterprise data.
130+ connectors covers basically every warehouse, BI tool, and pipeline system you'd actually encounter in a real data stack. Column-level lineage is genuinely hard to get right and they have it. The MCP server integration is practical — giving AI agents access to governed metadata rather than raw database connections is the correct architectural choice. Schema-first design with 700+ JSON Schemas means the metadata model is explicit and portable, not a proprietary blob.
Self-hosting is a serious operational commitment — the full stack requires Java, Python, Elasticsearch/OpenSearch, and a relational database, and the Docker Compose quickstart papers over real production complexity. The 'memory nuggets' feature sounds compelling in the README but the actual implementation appears to be glorified annotations attached to entities, not a queryable knowledge graph with inference. The README is written primarily as marketing for the 'AI context layer' narrative, which makes it harder to evaluate what actually works versus what's aspirational. MySQL and Postgres run through separate Flyway migration paths that need to stay in sync — that's ongoing maintenance debt that will bite teams running one database while the project primarily tests against the other.