---
title: "MDLH: Source syncing with bulk metadata"
url: "https://atlan.com/demos/mdlh-source-syncing-with-bulk-metadata/"
excerpt: "Break your metadata out of Atlan and into the AI agents and source systems that need it to work accurately."
description: "Make every month of enrichment effort actually travel. Pull asset descriptions, lineage paths, certification status, and ownership into a single export — then push it to Snowflake Cortex, Databricks Genie, a custom LLM agent, or any source system that needs Atlan as its authority. Set it up on a cadence and your entire downstream ecosystem stays synchronized with the context your organization has already built."
format: "Video"
duration: "PT3M25S"
video: "https://videos.ctfassets.net/nwa1c00rtgxb/66bEaUbjou71Yo2I5n0omn/6602e89ffe83924f8ec5ecf557005f5f/MDLH_-_Source_syncing_with_bulk_metadata.mp4"
thumbnail: "https://images.ctfassets.net/nwa1c00rtgxb/7heKzrybF5vM5INHhp7AuG/ea1fc7409f19252e440f5e5ed1ad5bd1/mdlh-source-syncing-with-bulk-metadata.webp"
content_purpose: ["Product Overview"]
target_persona: ["General", "Data Engineer", "Data Analyst", "IT Administrator"]
journey_stage: ["S2 - Discovery", "S3 - Solution Design", "S4 - Business Case", "C1 - Onboarding", "C2 - First Value", "C3 - Adoption"]
use_case_context: ["Training"]
product: ["Enterprise Data Graph - Integration"]
published: "2026-03-13"
updated: "2026-06-02"
content_type: "video transcript"
transcript_source: "sheet"
---

# MDLH: Source syncing with bulk metadata

Transcript of the video at https://atlan.com/demos/mdlh-source-syncing-with-bulk-metadata/

Your team has spent months enriching Atlan. But if that context never leaves Atlan, your AI agents are still working blind. One query changes that. There's a version of Atlan where all your enrichment effort stays inside Atlan. And a version where it actually travels — into AI systems, into source tools, into everything that needs it. Right now, most teams are in the first version. This query moves you to the second. We've done the work inside Atlan — our assets are described, our critical tables are documented, our teams have added context that makes data understandable.

But that context rarely makes it out. The moment someone needs it in a different system — an AI agent, a downstream tool, a compliance report — they're back to copying and pasting, exporting, or building automation through the APIs to get. The Metadata Lakehouse makes bulk metadata export straightforward — asset descriptions, lineage paths, and any business metadata in a single SQL query. Once we have the output, we can push it wherever it needs to go — giving AI agents and source systems the full context they need to work accurately. We can build a query against the Metadata Lakehouse to pull every asset that has a business description entered.

What we get back: every asset that has a description, its name for identification, its display name, and the type of asset it isHere's why this step matters more than it looks. Most AI context pipelines — whether you're using Snowflake Cortex, Databricks Genie, or a custom LLM agent — need asset descriptions to generate accurate answers. Without this export step, those agents are working blind. What you just ran is the query that fills that gap. Now let’s extend it — here’s how to pull in richer context alongside the descriptions. Descriptions are a great start, but we can pull more.

The Metadata Lakehouse also exposes owner, domain, certification status, tags, and lineage alongside each asset. Lineage is especially powerful here — it tells our downstream system not just what an asset is, but where its data comes from. An AI agent with lineage context can answer questions about data origin and trust, not just definitions. Adding certification status is also useful — it lets our downstream system distinguish between verified assets and works-in-progress. We don’t want an AI agent presenting an unverified definition as ground truth. Now let's look at what to do with this output once we have it.

Once we have this result set, we have two paths. We can export directly to CSV and load it into another system manually — that works for a first pass or a one-off push. Or we can connect a pipeline — Python, dbt, or a scheduler to run this query on a cadence so updates flow automatically. Either way, Atlan becomes the authoritative source for business metadata, and this query is how we distribute it. Set it up once on a cadence and Atlan's metadata stays synchronized with everything downstream. We now have a reusable query that pulls business context out of Atlan in bulk — descriptions, lineage, certification, owners, whatever our downstream systems need.

That lineage context is what lets AI agents and external systems understand not just what our data means, but where it comes from and whether it can be trusted. Set it up once, automate it, and our AI agents, our source tools, and our teams all stay in sync with the metadata our organization has already done the work to maintain. If we want to see how to track where that metadata coverage is strong and where it's missing — so you know which assets are ready to export and which ones still need work — that walkthrough is linked right here.

Drop a comment below if you’re building AI context pipelines — I'd like to know what systems you’re looking to build your context layer for.