Atlan MCP

Atlan's MCP server exposes the metadata catalog to an AI assistant. An agent can search assets by name, type, owner, classification or glossary term, then read the context Atlan holds on each one: description, certification status, owners, attached terms and classifications, and column level detail for tables and views. That gives a model the vocabulary of your data estate instead of a guess at table names. Lineage is the second capability. The server lets an agent traverse upstream and downstream from any asset, so a question like which dashboards break if a column is dropped becomes a graph walk rather than a manual click through the UI. Write tools cover curation: updating descriptions, certifying assets, assigning owners and attaching glossary terms, which is the tedious part of keeping a catalog trustworthy. Authentication uses an Atlan API token scoped to a workspace. It is aimed at data engineers, analytics engineers and governance teams already running Atlan. Typical uses are impact analysis before a schema or pipeline change, finding a certified table to build a new report on, and bulk documentation of assets that were imported without descriptions, driven from a coding assistant rather than the web interface.