Skip to main content

Best data catalogs for BigQuery and GCP

Emily Winks, Data Governance Expert, Atlan
Data Governance Expert
Updated:
|
Published:
26 min read

Key takeaways

  • Google renamed Dataplex Universal Catalog to Knowledge Catalog on April 10, 2026. It is the native BigQuery baseline.
  • Alation, Atlan, Collibra, DataHub, and OpenMetadata are compared only on what their own BigQuery docs state.
  • Knowledge Catalog is listed first as the native baseline. The other five follow alphabetically, with no scores.

What are the best data catalogs for BigQuery and GCP?

Google Knowledge Catalog, renamed from Dataplex Universal Catalog on April 10, 2026, is the native baseline: it crawls and indexes BigQuery datasets, tables, and views. Five third-party catalogs document BigQuery support in their own docs: Alation, Atlan, Collibra, DataHub, and OpenMetadata. They are listed alphabetically after the native baseline, not scored. The right pick depends on how much of your estate sits outside Google Cloud.

The six options compared:

  • Google Knowledge Catalog the native baseline, with automatic BigQuery entries and BigQuery lineage
  • Alation Google BigQuery OCF Connector, labelled a Core Connector in Alation docs
  • Atlan BigQuery down to columns and routines, query-history lineage, and a Knowledge Catalog connector
  • Collibra Knowledge Catalog inbound and outbound sync, plus GCP lineage in Collibra Data Lineage
  • DataHub BigQuery ingestion with optional table-level and column-level lineage
  • OpenMetadata BigQuery metadata, profiler, and usage workflows, with cross-project lineage

Not sure how ready your catalog stack is?


Six data catalogs fit BigQuery and GCP: Google Knowledge Catalog as the native baseline, then Alation, Atlan, Collibra, DataHub, and OpenMetadata in alphabetical order. Google renamed Knowledge Catalog from Dataplex Universal Catalog on April 10, 2026, and it crawls and indexes BigQuery datasets, tables, and views on its own. Google’s release note dated June 05, 2026 says the Data Catalog service begins a phased shutdown from June 1, 2026, and that Knowledge Catalog operates without impact.

Each of the five third-party catalogs documents BigQuery in its own docs, and none is scored. Knowledge Catalog can also pull in other sources through its connectors. For a data team feeding AI agents from BigQuery, the choice turns on where the rest of your metadata lives and which catalog the agents read from. Get it wrong and agents trace lineage that stops at the edge of Google Cloud, or read two different descriptions of the same table.

We read each vendor’s own BigQuery or Google Cloud documentation on November 17 and 18, 2026. Every product fact below comes from it; the advice on choosing is ours.

Build your BigQuery catalog shortlist


Give it the catalogs you are considering and every system around BigQuery. It returns an evaluation matrix covering connector coverage, where lineage stops, who maintains the metadata, and what happens if you leave. Read the skill.

Paste into a new chat

Use the skill at https://atlan.com/skills/catalog-tool-shortlist.md to evaluate a data catalog shortlist against our BigQuery and Google Cloud stack. Ask me for whatever it needs.

Run once in a terminal

curl -fsSL --create-dirs \
  -o ~/.agents/skills/catalog-tool-shortlist/SKILL.md \
  https://atlan.com/skills/catalog-tool-shortlist.md

For an agent

curl -fsSL https://atlan.com/skills/catalog-tool-shortlist.md

What counts as a data catalog for BigQuery and GCP?

A data catalog for BigQuery and GCP records what exists in your projects, from datasets and tables down to columns, plus who owns each asset and where its data came from, so people and AI agents can pick the right table. Metadata is data about your data: names, descriptions, owners, schemas, and lineage, the record of where data came from and what changed it.

Two shapes fit. The first is Google’s own Knowledge Catalog. According to Google’s metadata overview (Google Cloud, 2026), each project gets system-managed entry groups for Google Cloud services, such as @bigquery, and entries are “automatically populated from Google Cloud sources (such as BigQuery tables, views, datasets, and models).”

The second is a third-party catalog that reads BigQuery as one source among several. It does that through a connector: software that reads metadata from a source system and loads it into the catalog. Google documents the reverse direction too. To bring third-party metadata into Knowledge Catalog, you can use Knowledge Catalog connectors or a managed connectivity pipeline.

If AI agents will read the catalog, the questions shift from what exists to what it means: which definition applies, who owns it, and which source wins when two disagree. Data catalog vs context layer, what a context catalog adds, and context layer vs data catalog vs semantic layer set out how those jobs split.

For a BigQuery buyer, which shape fits depends on how much of the estate sits outside Google Cloud.


Best data catalogs for BigQuery and GCP: the comparison table

Google Knowledge Catalog is listed first because it is the native baseline for BigQuery. The five third-party catalogs follow in alphabetical order.

That order is Alation, Atlan, Collibra, DataHub, OpenMetadata. We do not score or rank them.

Every cell comes from the vendor’s own documentation. We did not read every vendor to the same depth, and the cells show it: “Docs not opened” means we did not open that vendor’s docs on the point, so the cell says nothing about what the product does. A row with more filled cells reflects which docs we opened, not a verdict on the product. Treat every “Docs not opened” cell as a question for that vendor.

The six options side by side


Catalog BigQuery assets in its docs Lineage in its docs Knowledge Catalog exchange Other Google Cloud services in its docs
Google Knowledge Catalog (native baseline, see the Knowledge Catalog guide) Crawls and indexes datasets, tables, and views; @bigquery entry group populated automatically For BigQuery, recorded for tables and views, within 24 hours of a job Native; imports third-party metadata through connectors or managed pipelines System entry groups for services such as Spanner and Pub/Sub
Alation Google BigQuery OCF Connector, labelled a Core Connector Query log ingestion, per a March 11, 2026 release note; lineage docs not opened Docs not opened Docs not opened
Atlan Projects, datasets, tables, views, materialized views, columns, and routines Column level, built by parsing query history Connector reads aspects, data quality, and profiling results; optional reverse sync Cloud Storage to BigQuery external table lineage, Cloud Composer, Dataflow, Looker
Collibra Entries in the BigQuery Entry Group, as Views, Entries, and Aspects Table or column level, through the Google Data Lineage API Inbound sync from Knowledge Catalog; outbound sync pushes enriched metadata back to aspects Cloud SQL (MySQL, PostgreSQL, SQL Server); lineage from other Google Cloud services into BigQuery
DataHub Projects, datasets, linked datasets; schema, descriptions, and usage by default Table level and column level, optional by configuration, from each project’s job history Docs not opened Column lineage with Cloud Storage, set by a separate option
OpenMetadata Metadata, profiler, and usage workflows; stored procedures and tags Cross-project lineage, with data ingested within a single service Docs not opened Docs not opened

How did we order these catalogs for BigQuery and GCP?

Knowledge Catalog comes first because it is the native baseline: Google maintains system entries for every BigQuery project without setup. The five third-party catalogs follow alphabetically.

No scores, no points, no “best for” column.

We fixed four rules before reading any vendor page:

  1. Use only each vendor’s own BigQuery or Google Cloud documentation, opened on November 17 and 18, 2026, plus Google’s dated release notes.
  2. Score nothing. The tables record what each vendor’s docs state.
  3. Give no product status (GA or Preview) unless a dated Google release note states it, and then quote Google’s word and date.
  4. Include only vendors whose docs we could open. Informatica is not included because its documentation pages returned HTTP 403 to our requests, so we could not read them.

Our reading was not even across vendors, and you should know where:

  • Google and Atlan: catalog, lineage, Knowledge Catalog, and agent (MCP) pages.
  • Collibra: the Knowledge Catalog integration page and the GCP lineage page.
  • DataHub: the BigQuery ingestion source page, including its lineage settings, and its MCP server page.
  • Alation: the BigQuery connector index and its release notes.
  • OpenMetadata: the BigQuery connector overview and its MCP server page.

Where we did not open a vendor’s docs on a point, the tables say “Docs not opened” and each profile says so. Treat those cells as questions for the vendor. The six options here are the ones we could document, not the whole market.

If you want to score your own shortlist, the data catalog RFP template and the 22 governance platform evaluation tests give you criteria to weight.


Google Knowledge Catalog: the native baseline

Knowledge Catalog is Google’s own catalog for BigQuery and the rest of Google Cloud. According to Google’s release notes (Google Cloud, 2026), Dataplex Universal Catalog was renamed Knowledge Catalog on April 10, 2026, and “the API, client library, CLI, and Identity and Access Management (IAM) names remain unchanged.”

Google’s Knowledge Catalog overview (Google Cloud, 2026) says it lets you “unify technical metadata and business semantics across structured and unstructured data using Gemini-powered capabilities.” For BigQuery, it automatically crawls and indexes datasets, tables, and views, and triggers Gemini-powered insights that analyze query histories, publish descriptions, and suggest verified example queries.

What the native catalog documents


  • Built-in entries. Each project has system-managed entry groups such as @bigquery, @spanner, and @pubsub.
  • Business glossary. Google describes glossaries, categories, and business terms that form a formal business taxonomy. Managing KPI definitions and a glossary in a data catalog covers the working side.
  • Third-party sources. Knowledge Catalog connectors or a managed connectivity pipeline bring external metadata in. Google’s own limitations note adds: “certain automated semantic extractions might be limited to built-in Google Cloud services.”
  • Lineage. For BigQuery, Google records lineage for tables and views, visible within 24 hours of a job and retained for 30 days.
  • Agent grounding. Google says you can expose metadata, schemas, lineage, and business rules to AI agents using local or remote MCP servers.

Recent dated entries in Google’s release notes, each with Google’s own status word:

  • November 12, 2026: data lineage supports Dataform workflows and BigQuery pipelines, generally available.
  • November 05, 2026: data products support more asset types, including Bigtable, Spanner, and BigQuery graphs, generally available. Packaging data products inside a catalog explains why that matters.
  • September 24, 2026: importing dbt metadata, generally available.
  • July 17, 2026: data lineage control at the organization, folder, or project level for BigQuery, generally available.

For how Knowledge Catalog compares with another platform’s native catalog, see Unity Catalog vs Dataplex. For who owns business meaning when Google’s stack and a neutral layer both hold it, read Google Agentic Data Cloud vs context layer and which layer is authoritative for BigQuery.

For an estate that lives mostly in Google Cloud, this baseline already covers a lot. The case for adding a catalog starts where your data leaves Google.


Alation: the Google BigQuery OCF connector

Alation documents BigQuery through its Google BigQuery OCF Connector, part of its Open Connector Framework. Alation’s connector page labels it a Core Connector and explains the term: “Core connectors are included with all Alation platform tiers (subject to each tier’s connector limits) and are fully supported by Alation.”

The newest entry in the release notes we opened is version 1.6.1, dated March 11, 2026. It fixed query log ingestion (QLI) parsing for queries with BEGIN, END, and MERGE syntax, which Alation notes also affected lineage extraction for those queries. The release history also lists per-user GCP service account certificate authentication for BigQuery data products.

One practical note for readers: Alation’s docs site says that from August 18, 2026, Alation Cloud Service documentation moved to a new site, and the site we read keeps Customer Managed documentation. Check which edition your team runs before you compare features.

We did not open Alation’s documentation on column-level lineage or on any Knowledge Catalog exchange, so those cells read “Docs not opened” in both tables.

If Alation already runs in your estate, the question becomes whether a second catalog should sit over it. Can a data catalog sit on top of existing catalogs without replacing them walks through that pattern.


Atlan: BigQuery crawling, query-history lineage, and Knowledge Catalog sync

Atlan is an AI Context Platform. It catalogs BigQuery as one source in a graph that also covers other warehouses and BI tools.

Crawling, connection, and previews


  • What it crawls. Projects map to Database and datasets to Schema, then tables, views, materialized views, columns, and routines. Table-valued functions become Function assets with column-level lineage flowing through them.
  • How it connects. Direct, over a public endpoint or Private Service Connect, or through a self-deployed runtime. Authentication uses a service account key or Workload Identity Federation, and Atlan uses those credentials to call BigQuery APIs for metadata only.
  • Previews read sample rows. Previews are a separate feature from crawling, and they do return data. Atlan runs no table scans: table previews use the BigQuery tabledata.list API, and for views and materialized views Atlan runs a query, shows a cost nudge first, and bills it to the project set in the miner’s billing strategy.

Lineage, Knowledge Catalog, and agents


  • Lineage. Atlan mines query history, with Direct or Agent extraction. The miner queries the past two weeks of history per run, a lookback window rather than a retention limit; for an initial load of older history, Atlan’s docs point to its Generic Miner. Warehouse lineage goes down to column level, and Looker appears in warehouse-to-BI lineage.
  • Knowledge Catalog exchange. Atlan’s Google Knowledge Catalog connector crawls aspect metadata, data quality results, and profiling results, and attaches them to the matching BigQuery assets. With reverse sync on, custom aspect changes in Atlan are written back; Atlan never writes Google’s system aspects.
  • Around BigQuery. Connectors for Cloud Composer (OpenLineage), Dataflow, Cloud Storage, and Looker, plus a workflow linking Cloud Storage objects to BigQuery external tables.
  • Agents. Through the Atlan MCP server, agents reach the context layer as 41 governed tools across nine categories, per Atlan’s MCP tools reference in November 2026. Write tools return a preview of the change and wait for approval before writing.

The same lineage docs name Databricks, Snowflake, and Amazon Redshift as warehouses Atlan builds lineage for alongside BigQuery; a multi-cloud context layer covers what changes when BigQuery is one platform of several. Data lineage for AI and metadata management for AI explain why agents need both.


Collibra: Knowledge Catalog ingestion and GCP lineage

The two Collibra pages we opened cover a Knowledge Catalog integration for metadata and Collibra Data Lineage for lineage. Both were dated early November 2026.

They describe BigQuery metadata arriving through Knowledge Catalog; we did not open Collibra docs on any other BigQuery ingestion path.

The Knowledge Catalog integration


Collibra’s page describes an inbound sync whose resulting assets “represent Views, Entries, and Aspects.” For BigQuery, “only assets that belong to the BigQuery Entry Group in Knowledge Catalog are ingested.” GCP Data Products and their ports can be brought in as an option. An outbound sync pushes enriched metadata back to aspects in Knowledge Catalog, and the same integration covers Cloud SQL for MySQL, PostgreSQL, and SQL Server.

Collibra states the scope plainly: the integration “ingests metadata only,” and technical lineage for GCP is set up through Collibra Data Lineage.

Collibra Data Lineage for GCP


  • Lineage for GCP “down to the column level,” with a choice of table-level or column-level lineage.
  • Built “via the Google Data Lineage API,” which needs the bigquery.jobs.get permission at minimum.
  • Lineage “can start from GCS or BigQuery, but can end only in BigQuery.”
  • In Collibra’s docs, GCS stands for other Google Cloud services. Collibra Data Lineage ingests the lineage those services generate for BigQuery assets.

Two-way sync with Knowledge Catalog suits an estate that wants both catalogs to hold the same business metadata; decide up front which system owns each field, so edits in one do not quietly overwrite the other. Unity Catalog vs Collibra shows how Collibra splits governance work with a platform’s native catalog.


DataHub: BigQuery ingestion in its own docs

DataHub documents BigQuery through an ingestion source with a capabilities table that says which features are on by default and which need configuration.

Enabled by default, per DataHub’s page:

  • Asset containers for projects, datasets, and linked datasets.
  • Dataset usage, descriptions, and schema metadata.

Optionally enabled by configuration:

  • Table-level lineage.
  • Column-level lineage.
  • Data profiling.
  • Domains.

The lineage settings carry detail worth reading before a pilot. With use_queries_v2 set to true, which is the default, “lineage and usage come from each ingested project’s own job history.” DataHub adds that if the jobs writing your tables run in a different project, such as a separate billing or backfill project, that project must be ingested too. Column lineage with Cloud Storage has its own option, include_column_lineage_with_gcs. A separate extract_lineage_from_catalog option reads Google’s Data Lineage API; for views, DataHub’s note recommends turning on view DDL parsing alongside it.

We did not open DataHub’s documentation on a Knowledge Catalog exchange, so we make no claim about it. Its MCP server is covered in the agents section below.

DataHub OpenLineage support covers its lineage model beyond BigQuery, and DataHub alternatives lists other options. If cost decides the shortlist, open source vs commercial data catalog total cost of ownership runs the numbers.


OpenMetadata: BigQuery connector and cross-project lineage

OpenMetadata documents BigQuery through a connector page in its docs.

What the connector page lists:

  • Metadata and profiler workflows, configured and scheduled from the OpenMetadata UI.
  • A usage workflow that reads INFORMATION_SCHEMA.JOBS_BY_PROJECT for usage data.
  • Data quality tests.
  • Stored procedures, ingested as their own assets.
  • Tags already defined on tables and columns in BigQuery.
  • dbt models and dbt lineage.

On lineage across projects, the page says: “OpenMetadata supports cross-project lineage, but the data must be ingested within a single service.” That shapes how you set up services for a multi-project BigQuery estate.

The connector page also lists lineage and column-level lineage among the connector’s available features. We did not open its lineage workflow pages or any Knowledge Catalog exchange, so we make no claim about them. Its MCP server is covered in the agents section below.

For how open lineage events from orchestration feed a catalog, see Airflow and OpenLineage for AI.


When is the native Knowledge Catalog enough for BigQuery?

Knowledge Catalog fits best when your estate lives mostly in BigQuery and other Google Cloud services. In that case, Google’s own docs cover the core catalog jobs:

  • BigQuery datasets, tables, and views indexed automatically, with Gemini-generated descriptions and example queries.
  • A business glossary with categories and terms.
  • Lineage for BigQuery tables and views, plus dated GA entries for Dataform, BigQuery pipelines, and Looker (Google Cloud core) lineage.
  • Agent access to metadata, schemas, lineage, and business rules through MCP servers.

Three conditions change the picture. The first two rest on Google’s own docs; the third is an ownership question no catalog settles for you.

Your sources reach past Google Cloud


Knowledge Catalog brings third-party metadata in through connectors or a managed connectivity pipeline. Google’s limitations note says “certain automated semantic extractions might be limited to built-in Google Cloud services.” If Snowflake, Databricks, or on-premises databases hold half your data, check which extractions you need on those sources. A quick test: list the 20 tables your most important reports and agents depend on, by platform. If several sit outside Google Cloud, test each candidate catalog on those first.

Your lineage questions span more than 30 days or leave Google Cloud


Google retains lineage for 30 days and, for BigQuery, records it for tables and views. Inside Google Cloud, its lineage docs also cover services such as Dataflow, Dataform, Cloud Data Fusion, and Looker (Google Cloud core); the gap opens at warehouses and BI tools outside Google. Write down the oldest lineage question your auditors have asked. If it reached back further than 30 days, ask each third-party vendor how long it keeps lineage it has built; none of their pages we opened stated that figure.

One set of business definitions must serve BigQuery and another platform


Knowledge Catalog’s glossary can hold those terms. The harder decision is ownership: when “active customer” is defined in BigQuery and again in another warehouse, someone has to say which definition an agent reads.

Extending a data catalog for AI readiness covers the extend-first route. Full-stack AI platform vs best-of-breed context layer frames the trade-off, and is your data catalog keeping up with what AI agents need tests the result.


How does BigQuery lineage compare across these catalogs?

BigQuery lineage differs across these catalogs in where it comes from, how deep it goes, and what limits each vendor documents. The table below quotes each vendor’s own docs.

Catalog Lineage source named in its docs Column level in its docs Limits stated in its docs
Google Knowledge Catalog BigQuery jobs; Data Lineage API with OpenLineage-format events for other systems Yes, for top-level columns in BigQuery tables Retained 30 days; no column-level lineage for load jobs, routines, nested STRUCT or JSON fields, upstream external tables, or jobs over 1,500 column links; custom OpenLineage events carry no column-level lineage
Alation Query log ingestion (QLI), per the March 11, 2026 release note Docs not opened Docs not opened
Atlan Parsed BigQuery query history, Direct or Agent extraction Yes, table-to-table and view-to-table, plus through table-valued functions Each mining run looks back two weeks of query history; Generic Miner for an initial load of older history
Collibra Google Data Lineage API, through Collibra Data Lineage Yes, table or column level by choice Needs bigquery.jobs.get at minimum; starts in BigQuery or another Google Cloud service and ends in BigQuery
DataHub Each project’s job history by default; Google’s Data Lineage API as an option Yes, optional by configuration For views, DataHub recommends view DDL parsing alongside the Data Lineage API extractor
OpenMetadata Ingested BigQuery services, plus dbt lineage Listed as an available feature on the connector page Cross-project lineage needs data ingested within a single service

Recent Google lineage entries


Two Google release-note entries extend the native column. On November 12, 2026, Google made lineage for Dataform workflows and BigQuery pipelines generally available. On October 26, 2026, Google made Looker (Google Cloud core) lineage generally available, covering LookML views, Explores, and dashboards.

Lineage that is parsed from SQL and lineage that is drawn by hand fail in different ways; automated SQL lineage vs manual lineage mapping shows where. Test it on your own jobs, the way a lineage proof of value does for Fabric. Retention and column coverage limit any answer that depends on lineage, because an agent can only trace what the catalog still holds; definition and table-choice questions need their own tests. MCP for data lineage shows how an agent asks.


Can AI agents use BigQuery metadata through these catalogs?

Yes. Four of the six document an MCP server in the pages we opened: Google Knowledge Catalog, Atlan, DataHub, and OpenMetadata.

MCP, the Model Context Protocol, is a standard way for an AI agent to call tools and read data from outside systems. Why MCP matters for AI agents and the MCP architecture deep dive cover the protocol.

Google Knowledge Catalog


Google’s overview says you can “expose metadata, schemas, lineage, and business rules directly to AI agents using local or remote MCP servers.” Two dated release-note items add to that:

  • November 02, 2026: Knowledge Studio, in Preview. It explores assets in natural language and an interactive context graph, runs enrichment tasks, and packages “context sets” that ground agents through MCP or the APIs.
  • June 04, 2026: the lookupContext method, in preview. It returns a pre-formatted, LLM-ready bundle of data asset context for agent workflows.

Atlan


The Atlan MCP server gives agents 41 governed tools across nine categories, including search, lineage, asset metadata, glossary, and data quality. Write tools return a preview and wait for approval before they change anything. Each preview shows a location breadcrumb for the affected assets, and a change that touches several fields arrives as one change set with a single approval. Atlan’s docs note that tools are gated per tenant, so a client may list fewer than 41.

DataHub and OpenMetadata


DataHub’s MCP server page says it gives AI agents access to DataHub metadata: search, schemas, and lineage at table and column level across multiple hops. It runs as a managed server on DataHub Cloud or self-hosted for DataHub Core, and its tools carry MCP-standard hints so clients can prompt before a tool modifies catalog state.

OpenMetadata documents an MCP server built into the platform that lets AI clients query metadata in natural language, with OAuth 2.0, personal access token, or bot token authentication.

Alation and Collibra


We did not open their MCP or agent documentation, so we make no claim either way. Ask each vendor for its current agent docs.

For the patterns behind agent access, read exposing AI-ready data to agents through MCP, how to connect Claude to a data catalog or context layer, MCP-connected data catalogs, and what AI agents for data catalogs can and can’t do.

The protocol is the easy part. Before you trust an agent’s answer, check which definitions, lineage, and owners its tools actually return for three of your own BigQuery tables.


What does a data catalog for BigQuery cost?

Knowledge Catalog bills pay-as-you-go, and third-party catalogs add their own BigQuery costs on top of a license. Budget for both.

According to Google’s Knowledge Catalog pricing page (Google Cloud, 2026), charges come from processing, metered in Data Compute Unit (DCU) hours at standard or premium rates, and from metadata storage. Data lineage, data quality, and data profiling run on premium processing. Data insights and enrichment tasks are billed in data tokens, at $3 per million input tokens and $20 per million output tokens. Some features, such as discovery scans and data profiling scans, may trigger jobs that are billed under BigQuery, Dataflow, or Cloud Storage instead of Knowledge Catalog.

Third-party catalogs read BigQuery too, and those reads land on a Google bill. Atlan’s view previews run queries billed to the project set in its miner’s billing strategy. OpenMetadata’s connector page has you set a billing project ID that tells BigQuery which project pays for its queries.

We did not open licensing or pricing pages for the five third-party catalogs. Ask each vendor for a quote sized to your project count, and ask which of its BigQuery reads you will pay for.


How do you choose a data catalog for a BigQuery estate?

Choose by the shape of your estate, then confirm each claim in the vendor’s own docs. Four common shapes:

  • BigQuery and Google Cloud only. Start with Knowledge Catalog. Its automatic BigQuery entries, glossary, lineage, and MCP access cover the core jobs. Revisit the choice if you need lineage history past Google’s 30 days or definitions shared across platforms.
  • BigQuery plus other clouds or warehouses. Decide whether you want one metadata graph across platforms or a sync between catalogs. Either way, check for column-level BigQuery lineage and a documented Knowledge Catalog exchange that carries the metadata you actually use. The same question shows up on AWS, covered in AWS Glue Data Catalog vs a cross-cloud governance layer and AWS Glue and DataZone vs a cross-cloud context layer, and on Azure in Purview data governance coverage.
  • An existing catalog already in place. Check how its BigQuery connector handles lineage and whether its docs describe a Knowledge Catalog sync. Collibra and Atlan both document an exchange.
  • AI agents as the main readers. Ask for MCP docs, the tool list, and how write actions are approved. Then test an agent question against your own BigQuery lineage.

For a scored version of this, use the RFP template above. Two sibling reads ask the same question on other platforms: best data catalogs for Microsoft Fabric and the best AI-ready data platforms.


Why the shortlist depends on what sits outside BigQuery

Knowledge Catalog already catalogs every BigQuery project, so a third-party catalog has to earn its place on something else: other warehouses and BI tools, lineage questions older than Google’s 30-day window (ask each vendor how long it keeps lineage), stewardship workflows your team already runs, or business definitions that more than one platform shares. Read the metadata management examples from eight programs to see how other teams drew that boundary.

For most BigQuery teams, the harder question comes later: when Knowledge Catalog and a second catalog describe the same table differently, which one do your AI agents read? Which layer is authoritative for BigQuery works through precedence rules for exactly that case.

If you want a baseline before the vendor calls, score your context readiness. The assessment gives you a maturity level, a score across six dimensions, a roadmap, and a PDF to share with your team.


FAQs about data catalogs for BigQuery and GCP

Eight questions BigQuery teams ask when they shortlist a catalog, answered from the same vendor docs.

1. What is Google Knowledge Catalog, and what happened to Dataplex Universal Catalog?


Knowledge Catalog is the new name for Dataplex Universal Catalog, as of April 10, 2026. Google says the API, client library, CLI, and IAM names stay the same. Per Google’s overview, it unifies technical metadata and business semantics across structured and unstructured data using Gemini-powered capabilities.

2. What is happening to Google Data Catalog?


Google’s release note dated June 05, 2026 says the Data Catalog service begins a phased shutdown from June 1, 2026, and that Knowledge Catalog operates without impact. From that date, Google warns of possible disruptions or a complete lack of access to Data Catalog APIs. Teams still calling those APIs should plan their move to Knowledge Catalog.

3. Does Knowledge Catalog catalog BigQuery automatically?


Yes. Google says Knowledge Catalog automatically crawls and indexes BigQuery datasets, tables, and views. Each project has a system-managed @bigquery entry group, populated from BigQuery tables, views, datasets, and models without manual setup.

4. How long does BigQuery lineage stay available in Knowledge Catalog?


Thirty days. Google’s lineage overview says all lineage information is retained for only 30 days, and that lineage appears within 24 hours after a BigQuery job completes. For BigQuery, Google records lineage only for tables and views.

5. How does Atlan connect to BigQuery?


Atlan connects directly, over a public endpoint or Private Service Connect, or through a self-deployed runtime. It authenticates with a service account key or Workload Identity Federation and calls BigQuery APIs for metadata only. Sample previews are a separate feature that returns rows: Atlan runs no table scans, table previews use the tabledata.list API, and view previews run a query with a cost nudge first.

6. Can a third-party catalog read Knowledge Catalog metadata?


Yes, two of the catalogs here document it. Atlan’s Knowledge Catalog connector reads aspect metadata, data quality results, and profiling results and attaches them to BigQuery assets, with optional reverse sync for custom aspects. Collibra’s integration syncs Knowledge Catalog entries in and pushes enriched metadata back to aspects.

7. How does Collibra bring BigQuery lineage in?


Through Collibra Data Lineage, which reads the Google Data Lineage API and can build table-level or column-level lineage for GCP. Collibra’s Knowledge Catalog integration ingests metadata only, so lineage runs as a separate setup. Collibra’s page says lineage can start from BigQuery or another Google Cloud service and ends in BigQuery.

8. How do DataHub and OpenMetadata ingest BigQuery lineage?


DataHub’s docs list table-level and column-level lineage as optionally enabled by configuration. In the default query mode, lineage and usage come from each ingested project’s own job history. OpenMetadata’s connector page lists lineage and column-level lineage, and says it supports cross-project lineage when the data is ingested within a single service.


Sources

  1. Knowledge Catalog release notes, Google Cloud
  2. Knowledge Catalog overview, Google Cloud
  3. Metadata management overview, Google Cloud
  4. About data lineage, Google Cloud
  5. Google BigQuery connector, Atlan documentation
  6. What does Atlan crawl from Google BigQuery, Atlan documentation
  7. How Atlan connects to Google BigQuery, Atlan documentation
  8. Mine Google BigQuery, Atlan documentation
  9. Routines and process assets, Atlan documentation
  10. Supported lineage types, Atlan documentation
  11. Google Knowledge Catalog connector, Atlan documentation
  12. GCS to BigQuery external table lineage, Atlan documentation
  13. Google Cloud Composer (OpenLineage) connector, Atlan documentation
  14. What does Atlan crawl from Looker, Atlan documentation
  15. MCP tools reference, Atlan documentation
  16. Google BigQuery OCF Connector, Alation documentation (docs.alation.com), accessed November 17, 2026
  17. Google BigQuery OCF Connector release notes, Alation documentation (docs.alation.com), accessed November 17, 2026
  18. About Dataplex Catalog ingestion, Collibra documentation (productresources.collibra.com), accessed November 17, 2026
  19. GCP supported transformations, Collibra Data Lineage documentation (productresources.collibra.com), accessed November 17, 2026
  20. BigQuery ingestion source, DataHub documentation (docs.datahub.com), accessed November 17, 2026
  21. BigQuery connector, OpenMetadata documentation (docs.open-metadata.org), accessed November 17, 2026
  22. DataHub MCP Server, DataHub documentation (docs.datahub.com), accessed November 18, 2026
  23. MCP Server, OpenMetadata documentation (docs.open-metadata.org), accessed November 18, 2026
  24. Google Dataflow connector, Atlan documentation
  25. Knowledge Catalog pricing, Google Cloud

Share this article

signoff-panel-logo

Atlan is the Context Layer for AI. It translates business knowledge, including data definitions, working procedures, and governance policies, into context AI can actually use. This knowledge lives in a single Enterprise Data Graph that every team and AI agent can reach.

In Atlan's research report, rich semantic metadata improved the accuracy of AI-generated SQL by 38%.

Atlan is recognized as a Leader across multiple Gartner reports and Forrester Waves, and is trusted by over 400 enterprises representing $10T+ in market cap, including Mastercard, Workday, General Motors, CME Group, HubSpot, FOX, Virgin Media O2, and Elastic.

Bridge the context gap.
Ship AI that works.