What are the best data catalog tools in 2026?
Seven tools ranked so your teams find the right data, understand where it comes from and trust it, with what each one really costs.
The answer depends first on where your data lives. If everything runs on Databricks, Unity Catalog is already included. If everything sits in Azure and Power BI, Purview plugs in without a project. In every other case you need a dedicated catalog, and the choice comes down to three profiles.
For a data team on a cloud warehouse, Atlan is our top pick: the interface most built for business users, with AI agents and an MCP server. To start with zero budget, OpenMetadata and its free Collate tier. If you want a European vendor with a strong business glossary, DataGalaxy. Regulated enterprises look at Collibra, platform teams at DataHub. One thing almost all of them share: no public price. Below, you'll see exactly who shows what.
Some links may be affiliate links. The ranking is editorial: a sponsored spot is always labeled on its card, and no affiliate link changes a score.
The best data catalog for your case
Built for business users, AI agents and MCP
Free startCollate Free: 5 users, 500 assets, $0
European vendorStrong business glossary, EU or US hosting on Portfolio
Regulated enterpriseThe deepest governance, at a premium price
Already on DatabricksIncluded, no separate license
All-in on MicrosoftUsage-based, billed per governed asset
The 7 data catalog tools compared
Read the "What it covers" column first, then the price. A catalog does three jobs: help people find data (search and descriptions), understand it (glossary and lineage, meaning the path data takes from its source to the dashboard) and govern it (owners, access, sensitive data). Each tool leans on a different one.
| Best for | What it covers | Free plan | Team size | Visit | ||||
|---|---|---|---|---|---|---|---|---|
| 1 | Atlan | Modern data team | Catalog, glossary, lineage, AI agents, MCP | 4.2/5 | Quote only | — | Data team of 5 to 50 | Visit → |
| 2 | OpenMetadata (Collate) | Free start | 130+ data services, glossary, MCP, self-hostable | 4.1/5 | $0 (Collate Free) | ✓ | Small team, zero budget | Visit → |
| 3 | DataGalaxy | European vendor | Catalog, glossary, lineage, use-case portfolio | 4.1/5 | Quote only | — | Mid-size and large companies | Visit → |
| 4 | Collibra | Regulated enterprise | Governance, sensitive data classification, data contracts | 4.0/5 | Price on request | — | Enterprise with a compliance team | Visit → |
| 5 | DataHub | Open source, large team | Column-level lineage, versioned glossary, data quality | 4.0/5 | Free (self-hosted) | ✓ | Data platform team | Visit → |
| 6 | Databricks Unity Catalog | All-in on Databricks | Automatic column lineage, access control, AI models | 3.9/5 | Included (Premium, Enterprise) | — | Databricks customer | Visit → |
| 7 | Microsoft Purview (Unified Catalog) | All-in on Microsoft | Unified Catalog, Data Map, classification | 3.8/5 | Usage-based, per asset | — | Azure, Fabric, Power BI customer | Visit → |
Prices read on each vendor's official pages in 2026. "Quote only" means the official page offers nothing but a form or a demo request.
Is your tool missing from this page? Email us: hello@hackceleration.com
How we compared and ranked these tools
We put ourselves in the shoes of a head of data or a CIO at a company of 100 to 5,000 people. Their question isn't "which tool has the most features?" It's "which one will my teams actually open, and what will it cost me?" We reviewed sixteen catalogs, read their official product and pricing pages, and checked what they promise against what data engineers say on their forums and in user reviews.
The pages that rank on this query today are mostly written by vendors who put themselves first, or by software directories with no opinion. Several still list tools that have been acquired and are no longer sold on their own. Almost none tell you who publishes a price. We do, and we tell you who each tool is not for.
Our entry rule: a tool gets ranked only if its official page describes at least a glossary or lineage, and if we could read its price or billing model on the vendor's site. Acquired tools, and tools whose facts we couldn't read on an official page, sit in the "Not ranked" section instead.
- Business coverageLineage (ideally down to the column), glossary, governance and sensitive data. It's the reason you buy a catalog in the first place.30%
- Adoption by business teamsA catalog only engineers open will die. Interface, search, generated descriptions.20%
- Cost and price transparencyPublic price or quote, free tier, and the running cost of a self-hosted tool.20%
- Connectors and ecosystemHow many sources plug in without custom code, and how well.15%
- Sovereignty and hostingControl over where your metadata lives: region choice, single-tenant or self-hosted options.15%
Some links may be affiliate links. The ranking is editorial: a sponsored spot is always labeled on its card, and no affiliate link changes a score.
Atlan
Atlan is the catalog that data teams running Snowflake, dbt and a BI tool mention most often. It puts the catalog, the business glossary, lineage, governance and data quality in one interface. It lists 110 native connectors, and it can sit on top of catalogs you already run, such as Microsoft Purview or Snowflake Horizon, pulling in their metadata. That matters if your company already has a native catalog in one cloud and wants a shared layer on top.
Its real edge in 2026 is AI. Agents scan your data estate, draft table descriptions, link glossary terms to columns and suggest metrics. An MCP server (the protocol that lets an assistant like Claude or ChatGPT query a tool) exposes that context to your own AI agents. In practice, your assistant knows what "net revenue" means at your company before it writes the SQL.
The catch is price. There's no plan and no number on the site, just a meeting request. User reviews also mention a steep learning curve for newcomers, and third-party connectors outside the big warehouses that still need work.
- 110 native connectors, including Snowflake, dbt and Airflow
- AI agents that draft descriptions and link the glossary
- MCP server to feed context to your AI agents
- Sits on top of Purview and Snowflake Horizon
- ✓The most business-friendly interface in this ranking
- ✓Glossary, lineage and governance in a single tool
- ✓A shared layer across several clouds
- ✗No public price: you go through a sales call
- ✗Steep learning curve for new users
- ✗Hosting options aren't spelled out on the site: ask in the demo
The best pick if your warehouse lives in the cloud and you want marketing or finance to find the right table on their own. Show up to the demo with your list of sources and ask for a price based on your user count, not a vague package. If your budget can't absorb a five-figure annual subscription, look at OpenMetadata first.
OpenMetadata (Collate)
OpenMetadata is the open source catalog that comes up most when people look for a free option. It covers 130+ data services, a glossary (terms, metric definitions, semantics) and an MCP server to plug in your AI agents. You can install it on your own infrastructure, so your metadata lives wherever you decide.
If you don't want to run anything, Collate, the company behind the project, offers a free hosted tier. Its limits are clear: 5 users, 500 data assets, a shared cluster, unlimited connectors but a weekly refresh, and no SSO. One trap to know about: an inactive cluster gets reclaimed after 4 weeks. Premium (25 users, 5,000 assets) and Enterprise (50 users, 10,000 assets, unlimited as an option) go through a demo, with single-tenant SaaS, hybrid SaaS or a deployment in your own cloud.
On data engineering forums, OpenMetadata is often called the best open source bet, with one reservation: it's still more limited than a high-end commercial catalog. It also needs fewer moving parts than DataHub, which means less maintenance.
- 130+ data services covered
- Collate Free at $0: 5 users, 500 assets
- MCP server for AI agents
- Self-hostable: you choose where your metadata lives
- ✓The only true free start in this ranking
- ✓Free-tier limits written out in plain terms
- ✓Lighter to run than DataHub
- ✗Collate Free: weekly refresh, no SSO
- ✗Free cluster reclaimed after 4 weeks of inactivity
- ✗Less polished for business users than Atlan or DataGalaxy
The best starting point if you need to prove a catalog is worth it before asking for budget. Start with Collate Free on one source, show the result to two or three teams, then decide: self-host, Premium, or a commercial tool. Budget engineering time if you host it yourself.
DataGalaxy
DataGalaxy is a French data and AI governance vendor, and the one most US comparisons skip. It sells two products. DataGalaxy Catalog covers data discovery, the business glossary, lineage, governance and quality. DataGalaxy Portfolio manages your use cases, requests and data products, with tracking of the value they create. Listed integrations include Snowflake, Databricks, Power BI and Looker, among others.
On price, nothing is displayed: every plan points to a quote request, and no free trial is mentioned. On hosting, Portfolio Start offers SaaS in Europe or the United States, and Scale a dedicated tenant in the region of your choice. For the Catalog product, the pricing page doesn't state the region: get it written into the contract.
Public reviews are very strong, around 4.8 out of 5, but light on detail. That's a good sign, not proof: ask for two reference customers your size and call them.
- European vendor, a natural fit for GDPR-minded buyers
- Catalog and use-case portfolio in the same suite
- EU or US hosting stated on Portfolio Start
- Integrations with Snowflake, Databricks, Power BI, Looker
- ✓Glossary and governance that speak the business language
- ✓Use-case value tracking, rare among competitors
- ✓Strong option for a company with teams on both sides of the Atlantic
- ✗No public price, no free trial
- ✗Hosting region for the Catalog product must be clarified
- ✗Fewer listed connectors than Atlan or Collibra
The natural choice if your leadership wants a European vendor and a tool business teams will adopt, or if you run data in both the EU and the US. Get the Catalog hosting region and the exact connector list in writing. If your stack is heavily dbt, Snowflake and Airflow, run Atlan on the same sources before you sign.
Collibra
Collibra is the tool large organizations use when they must show who owns which data, who can access it and where it comes from. Its catalog organizes lineage, classifications, owners and policies, with interactive diagrams to follow data flows. A semantic layer ties technical data to glossary terms and policies. Automatic classification flags sensitive categories such as personal data (PII) and health data (PHI).
It lists 100+ native connectors (clouds, databases, applications, BI tools, legacy systems), an AI assistant that answers questions about your data and drafts descriptions, plus data contracts and an internal marketplace to publish data products.
The downside is well known: it's one of the most expensive options, and user reviews describe modules sold separately (governance, lineage, quality). Reported rollouts range from a few weeks to several months. No price is published.
- 100+ native connectors, legacy systems included
- Automatic classification of PII and PHI
- Data contracts and an internal data marketplace
- AI assistant to query the catalog
- ✓The most complete governance in this ranking
- ✓Built for audits and regulators
- ✓Covers old on-prem systems, not just the cloud
- ✗Among the priciest, with modules billed separately
- ✗Long rollout: from a few weeks to several months
- ✗Too heavy for a small company or a data team under 10
The pick for organizations with a compliance team and a regulator to satisfy. In the demo, ask for pricing module by module (catalog, lineage, quality) and a written rollout plan. For a smaller team, Atlan or DataGalaxy will get you there faster.
DataHub
DataHub was born at LinkedIn and remains one of the most used open source catalogs. It ships under the Apache 2.0 license, so you can install and modify it freely. It traces lineage at the column level, from source to dashboard, offers a business glossary with owners and version history, and adds data quality and freshness checks.
The trade-off is operations. DataHub needs more components than OpenMetadata (a search engine, a message queue, a graph store), and each one has to be upgraded, monitored and fixed when it breaks. As a rough order of magnitude, a self-hosted catalog eats between a quarter and a full platform engineer. If you don't want to carry that, DataHub Cloud is the vendor-managed version, with a service commitment.
Data engineers in community threads highlight its usage history on BigQuery and Snowflake: you see which tables are actually queried, which is the fastest way to find dead tables.
- Apache 2.0 license, free and modifiable
- Column-level lineage
- Glossary with owners and versions
- Data quality and freshness checks
- ✓The most mature column lineage among free tools
- ✓Large community, lots of integrations
- ✓A managed version exists if you want to stop hosting
- ✗Heavy to run: several components to maintain
- ✗Interface less welcoming for business users
- ✗DataHub Cloud pricing isn't published
The right pick if you have a platform team and column lineage matters more to you than the interface. Price out the engineering time first, or "free" becomes your most expensive line item. If you're a team of one or two, start with OpenMetadata.
Databricks Unity Catalog
If you're already on Databricks, the question isn't "which catalog should I buy?" It's "do I need anything beyond Unity Catalog?" It's included in the Premium and Enterprise plans: no separate license, you only pay for the compute and storage you already use. It traces lineage automatically down to the column, for your tables and for your AI models.
Teams that move to Databricks often adopt it by default, and that makes sense: the catalog already knows everything about your tables and access rights. The limit shows up as soon as data leaves Databricks. A CRM, a finance spreadsheet or a legacy database stay on the sidelines, and lineage outside the platform is weak.
- No separate license on Premium and Enterprise
- Automatic lineage down to the column
- Also governs AI models
- Access control managed where the data lives
- ✓The lowest marginal cost if you're already a customer
- ✓Nothing to install or sync
- ✓Reliable lineage, computed by the platform itself
- ✗Built for Databricks, weak on the rest of your estate
- ✗Less glossary and business context than a dedicated catalog
- ✗Requires a Premium or Enterprise plan
Start here if you're on Databricks, and only buy a dedicated catalog the day business users can't find their way, or half your data lives elsewhere. Atlan can then read its metadata without starting from scratch.
Microsoft Purview (Unified Catalog)
Microsoft Purview is the obvious catalog if your data lives in Azure, Fabric and Power BI: it connects without an integration project, and it inherits your accounts and permissions. Its newer catalog (Unified Catalog) is billed on usage, per governed asset. The first 1 MB of Data Map storage is free for every customer.
For a ballpark, Microsoft publishes an example: 200 governed assets cost $100 over 30 days, without discounts. Divide, and you get about $0.50 per asset for the example period. That's our math from the example, not a listed rate: rerun it in the Azure pricing calculator with your asset count and region before you budget.
The limits users report are clear: gaps as soon as you connect technologies outside the Microsoft stack, and navigation that business teams find less effective. The new catalog also requires moving to the enterprise version.
- Plugged into Azure, Fabric and Power BI out of the box
- Usage-based billing, no per-user subscription
- First 1 MB of Data Map free
- Data classification in the Data Map
- ✓Fast start if everything is already on Microsoft
- ✓Cost follows your number of governed assets
- ✓Users rate setup time and integration well
- ✗Weak on non-Microsoft sources
- ✗Navigation rated less effective by business users
- ✗Hard to forecast the bill without the calculator
The right pick if 80% of your data sits in the Microsoft world. Start by governing a hundred critical assets, check the first month's bill, then expand. If a big share of your data lives in another cloud, Atlan or Collibra on top of Purview will hold up better.
Which data catalogs actually publish a price?
It's the question every buyer asks and no ranking answers. Out of sixteen catalogs we reviewed, not one dedicated commercial catalog shows a public price. The only real numbers come from the clouds and from Collate's free tier. Here's what you can know before your first demo.
Free: OpenMetadata, DataHub and Collate Free
$0OpenMetadata and DataHub are open source: the software is free, the hosting and the engineering time are not. Collate Free is the only free hosted option: 5 users, 500 data assets, weekly refresh, no SSO. A cluster left idle for 4 weeks gets reclaimed, so don't forget about it over the holidays.
The tip: 500 assets is one or two mid-size databases. Enough for a pilot, not for your whole warehouse.
- Free hosted
- Collate Free
- Limit
- 5 users, 500 assets
Usage-based: Purview and Amazon DataZone
PricedPurview bills its catalog per governed asset; the official example lands at $100 for 200 assets over 30 days, without discounts. Amazon DataZone publishes its full list: $10 per 100,000 requests (4,000 free each month), $0.40 per GB of metadata (20 MB free), $1.776 per compute unit, with no upfront fees or long-term commitment.
The tip: these prices look small, but they climb with assets and requests. Run one real month before you roll out everywhere.
- DataZone
- $10 per 100,000 requests
- Purview
- Per governed asset
Included: Unity Catalog
No licenseUnity Catalog needs no separate license on Databricks Premium and Enterprise. You pay for the compute and storage you already consume. For a Databricks customer, it's the lowest marginal cost on the market.
- Condition
- Databricks Premium or Enterprise
Quote only: Atlan, Alation, Collibra, DataGalaxy and the rest
DemoAtlan shows a meeting form. Alation offers a 30-minute call. DataGalaxy lists its plans (Catalog Starter, Professional, Enterprise; Portfolio Start, Scale) but sends each one to a quote. Collibra and Zeenea publish no price list. The six-figure numbers you'll find online come from competitors or procurement brokers and contradict each other: don't anchor on them.
The tip: send the same one-page brief to three vendors (number of sources, assets, active editors and read-only viewers) so the quotes are comparable. Then ask each one what happens to the price in year two.
- Tools affected
- Most dedicated catalogs
Pick your catalog based on where your data lives
Before comparing features, look at your warehouse. A platform's automatic lineage is reliable inside that platform and turns manual the moment you step outside it. That's your first filter.
You're on Databricks
Unity CatalogTurn on Unity Catalog and document your critical tables before you buy anything. Add a dedicated catalog only if business users can't find their own vocabulary in it, or if a big share of your data lives outside Databricks.
You're on Microsoft (Azure, Fabric, Power BI)
PurviewPurview connects without a project and bills on usage. Watch out for non-Microsoft sources (Salesforce, a warehouse on another cloud): that's where users report gaps.
You're on AWS
DataZoneAmazon DataZone is the native catalog, with a public price list and a free monthly allowance. It sits in the "Not ranked" section below: a good add-on to an AWS account, less built for an estate spread across several clouds.
You're on Snowflake
Horizon firstSnowflake bought Select Star, so check what the platform's own Horizon features already give you before you shop. If you need a business glossary and lineage across dbt and your BI tool, Atlan reads Horizon metadata and sits on top.
Your data is everywhere (warehouse, dbt, SaaS apps, legacy databases)
Dedicated catalogThis is where a dedicated catalog really wins. Atlan if you want business adoption and AI, DataGalaxy if a European vendor matters, Collibra if a regulator audits you, OpenMetadata or DataHub if you have engineers and no license budget.
What kills a data catalog, and how to avoid it
On data engineering forums, the top complaint isn't technical. It's "nobody uses it": people end up pinging the data team anyway, especially when the catalog looks complicated. Here's what keeps coming up, and what works.
Name an owner before you buy
A catalog nobody maintains goes stale in three months, and then nobody trusts it. Assign one owner per domain (finance, marketing, product), with one hour a week blocked in their calendar.
Start with 50 tables, not everything
Document the tables that feed your leadership dashboards first. That's where the questions come from. The rest follows once people get the habit.
Use the free lineage in your warehouse
BigQuery already captures lineage automatically, and so does Unity Catalog. Plug those in before you pay for lineage elsewhere. Just know that for sources outside the platform, you'll often add lineage by hand.
Let AI draft, then have a human review
AI-written descriptions (Atlan, Collibra) save weeks. But a wrong description does more damage than an empty one: have the domain owner approve them before they go live.
Count viewers, not just editors
Pricing often depends on user count. Split the people who document (often fewer than 20) from the people who read (hundreds). Some vendors charge little or nothing for viewers, and Secoda lists unlimited viewers on every plan: ask for it in the demo.
Watch out for acquisitions
Select Star joined Snowflake, Secoda was bought by Atlassian, data.world by ServiceNow, Informatica by Salesforce. Ask your vendor for a three-year roadmap and a metadata export clause in the contract.
- List of your sources and your main platform
- Number of assets to catalog in year one
- Editors and viewers counted separately
- Hosting region written into the contract
- Metadata export if you leave
- A named owner for each domain
Other data catalogs: good tools, not ranked
These tools show up in every search. They aren't ranked either because they were acquired, or because we couldn't read enough official facts to score them fairly.
Alation: the catalog pioneer
On requestAlation is the name many US enterprise buyers start with. It lists 120+ prebuilt connectors, an open framework to build more, and end-to-end lineage. Its pricing page offers nothing but a 30-minute call. Put it on your demo shortlist if you're comparing Atlan and Collibra, and ask early how its licenses are split between editors and viewers.
- Connectors
- 120+
- Price
- On request (sales call)
Zeenea (Actian): the French-born catalog
DemoZeenea, founded in France and acquired by Actian, promises a rollout in days and 100+ connectors. No price is published, only a demo. We couldn't read enough functional detail to score it.
- Connectors
- 100+
Amazon DataZone: AWS's native catalog
Usage-basedPublic price list, no per-user subscription: $10 per 100,000 requests with 4,000 free each month, $0.40 per GB of metadata with 20 MB free. AI recommendations cost $0.015 per 1,000 input tokens and $0.075 per 1,000 output tokens. Worth a look if everything runs on AWS.
- Free
- 4,000 requests a month
Google Dataplex: for BigQuery shops
Google CloudGoogle Cloud's catalog, worth checking if your warehouse is BigQuery, which already captures lineage automatically. We couldn't read its price list, which is why it isn't ranked: check it on the Google Cloud pricing page.
- For
- BigQuery customers
Secoda: acquired by Atlassian
AcquiredSecoda joined Atlassian in 2025. Its plans (Core, Premium, Enterprise) still show no price, with unlimited viewers on every plan and table and column lineage from Core. A good fit for a small team, worth watching until its place inside Atlassian is clear.
- Viewers
- Unlimited
Select Star: no longer sold on its own
SnowflakeSelect Star is now part of Snowflake and no longer shows a price. If you're on Snowflake, see what the platform already includes before you look elsewhere.
- Status
- Part of Snowflake
Pick your next step based on what you need
A catalog doesn't do everything. Here are our related rankings, and the parts of this page that answer the most common questions.
Dashboards your teams will read
You want to document processes, not dataBest knowledge base softwareInternal wikis compared
You want to analyze how people use your productBest product analytics toolsProduct analytics compared
You want to automate your data flowsn8n reviewThe self-hostable automation tool
You're building the budgetWho shows a priceFree, usage, included, quote
You're worried nobody will use itField tipsWhat kills a data catalog
Data catalog tools: frequently asked questions
What is a data catalog tool?
It's a searchable inventory of your data: every table, dashboard and pipeline, with a description, an owner, where it comes from and who can access it. It answers one question: which table should I use for this number, and can I trust it? Most tools pull metadata automatically from your warehouse, databases and BI tools, so you don't type it all in by hand.
What's the difference between a data catalog and a data dictionary?
A data dictionary describes the columns of one database: name, type, definition. A catalog covers all your sources and adds a business glossary, lineage, owners and access rights. A well-kept dictionary in a spreadsheet is often the smartest first step before you buy a catalog, because it shows you which 50 tables actually matter.
Is a data catalog the same as metadata management?
Not quite. Metadata management is the discipline: collecting, storing and governing information about your data. The catalog is the tool people search, the front door to that metadata. In practice the modern tools on this page do both, which is why vendors now pitch them as metadata or governance platforms.
Is there a free or open source data catalog?
Yes. OpenMetadata and DataHub are open source and free to self-host. If you don't want to run anything, Collate Free hosts OpenMetadata for $0 with 5 users and 500 data assets. The real cost of free is engineering time: as a rough order of magnitude, a self-hosted catalog takes between a quarter and a full platform engineer.
How much does a data catalog cost?
Almost no vendor will tell you upfront. Atlan, Collibra, DataGalaxy and Alation are all quote-only. The only published numbers come from the clouds: Amazon DataZone shows a full usage-based price list, and Purview bills per governed asset ($100 for 200 assets over 30 days in Microsoft's own example). Unity Catalog is included in Databricks Premium and Enterprise.
How long does it take to implement a data catalog?
A few days to connect your first sources, several months to get real adoption. User reports on Collibra range from a few weeks to several months of rollout. The slow part isn't technical: it's documenting the tables that matter and getting a named owner for each data domain.
Should I pick an open source or a commercial data catalog?
Pick open source if you have platform engineers and no license budget, and you care more about lineage than about the interface. Pick commercial if business users must adopt it without training, or if you need audit-ready governance. A hosted open source tier like Collate Free is a cheap way to find out which side you're on.
Do you need a data catalog for AI agents and LLMs?
More and more, yes. An agent writing SQL needs to know which table is the source of truth and what each metric means at your company. Atlan and OpenMetadata expose that context to your agents through an MCP server, and DataHub exposes it too. Without it, the agent guesses, and gets your business definitions wrong.
Which data catalog is best for a small company?
A company under 200 people rarely needs Collibra. Start with the catalog your platform already includes (Unity Catalog, Purview) or with Collate Free. Move to a commercial tool the day several business teams ask the data team the same questions every week.
