OpenMetadata
OpenMetadata's built-in MCP server lets an AI assistant query lineage, ownership, and glossary metadata across your data stack in plain language.
Install from the in-app marketplace
<your-openmetadata-server>/marketplace/apps/McpApplication → Install
About
OpenMetadata ships its own MCP server as an installable in-platform application rather than a separate process. You install it from the marketplace (Marketplace → Apps → McpApplication) and it exposes an /mcp endpoint that reuses OpenMetadata's existing role-based access control, so an agent only ever sees what the querying user is already allowed to see. The point is letting a model answer questions grounded in your actual metadata graph (lineage, ownership, glossary terms, freshness) instead of guessing.
Once installed, it connects to Claude Desktop, Claude Code, Cursor, VS Code, ChatGPT, or Goose over OAuth 2.0 (reusing your existing SSO login across Google, Azure, Okta, Auth0, LDAP, or SAML) or a personal access token for headless setups. Typical use is asking where a metric comes from, which pipeline last touched a table, or what's connected upstream and downstream of a dashboard.
Key features
- Installs as an in-platform app rather than a standalone server process
- Reuses OpenMetadata's existing role-based access control for every MCP request
- OAuth 2.0 login (Google, Azure, Okta, Auth0, LDAP, SAML) or personal access tokens
- Natural-language lineage tracing across tables, pipelines, and dashboards
- Semantic search over the metadata knowledge graph
- Works with Claude Desktop, Claude Code, Cursor, VS Code, ChatGPT, and Goose
Use cases
- Asking an assistant which pipeline last updated a given table and when
- Tracing a dashboard's full upstream lineage before making a schema change
- Drafting glossary term definitions from existing metadata context
- Letting non-technical stakeholders query data ownership and freshness in plain English
Available tools
Metadata Q&A
Answers natural-language questions about dataset ownership, definitions, and update timestamps.
Lineage tracing
Follows how data flows through pipelines and dashboards, upstream and downstream.
Semantic search
Finds relevant metadata assets by meaning rather than exact keyword match.
Knowledge graph access
Queries the relationships OpenMetadata has built across all connected data assets.
Frequently asked questions
Is this a separate server I have to run?
No. It's an application you install from inside OpenMetadata itself; installing it turns on an /mcp endpoint on your existing server.
Can it expose data a user shouldn't see?
No. Every MCP request goes through the same role-based permissions as the regular OpenMetadata UI and API.