Atlan is the data catalog that other data catalogs now copy. Founded in 2019 by Prukalpa Sankar and Varun Banka — both data engineers who had spent the previous six years running SocialCops, an Indian data consultancy that built analytics for the UN, WHO, Gates Foundation, and the Indian government — Atlan was built out of a very specific frustration: the founders had tried every existing catalog on actual, large-scale projects and hated all of them.

The origin story matters because it explains the product. Prukalpa and Varun were not outsiders theorizing about metadata; they had personally managed data teams working across hundreds of datasets for government clients and found that the existing catalog tools (Collibra, Alation, Informatica) were designed for governance committees, not for the engineers and analysts who actually touch the data. Their bet was that a catalog should feel like a modern SaaS product that a data team voluntarily opens — like Notion, Figma, or Linear — rather than a compliance system people are forced into.

That bet is now the dominant assumption in the category. In 2026, almost every modern catalog pitch sounds like Atlan's 2019 pitch.

What Atlan Actually Does

At the core, Atlan is the same four things every catalog is: inventory, lineage, search, and governance. What distinguishes it is the execution and the opinions baked into the defaults.

The collaboration layer. Atlan treats a data asset the way Figma treats a design file: something teams comment on, tag each other in, certify, and discuss inline. Every table has a description, an owner, a domain, a readme, and a conversation thread. Slack and Jira integrations push catalog activity into the tools teams already live in. The founders frequently describe Atlan as a "collaboration workspace for data teams," and the product really does behave more like Notion than like a legacy metadata repository.

Deep dbt and Snowflake integration. Atlan made an early bet that dbt would become the standard transformation layer — a bet that turned out to be correct — and the product is built around it. Atlan natively ingests dbt models, tests, exposures, and docs; column-level lineage is derived from dbt's compiled SQL; and dbt metadata becomes first-class inside the catalog. The same depth exists for Snowflake, BigQuery, Databricks, Looker, Tableau, and Power BI. If your stack is "modern data stack classic," Atlan fits like a glove.

Column-level lineage out of the box. Atlan ships automated column-level lineage across SQL warehouses and dbt, parsed from query logs and transformation code. This lets a data engineer click any column — say revenue in a dashboard — and walk backwards through every transformation, join, and filter that produced it. Column-level lineage is the single most important discriminator between modern and legacy catalogs, and Atlan was one of the first vendors to make it feel polished.

Active metadata and the "control plane" pitch. Atlan uses the term active metadata heavily: the idea that metadata shouldn't just sit there but should drive automation — tagging PII downstream, pushing alerts when upstream schemas change, propagating owners, triggering dbt reruns. In practice this is a set of workflows and webhooks, but the framing has been influential; Gartner adopted similar language in its market guides.

A design sensibility that is, frankly, a competitive moat. Worth saying plainly because no spec sheet captures it: Atlan is the best-looking product in its category by a noticeable margin. In a market where the alternative is UIs designed in 2012, "pretty" is a real feature because analysts will actually open it.

Architecture

Atlan is built on top of Apache Atlas, the open-source metadata framework originally created at Hortonworks, which the team forked and heavily extended. Underneath, it uses JanusGraph for the lineage graph, Elasticsearch for search, and a microservices backend for ingestion and policy. The product is delivered as a multi-tenant SaaS with a single-tenant deployment option for regulated customers. Crawlers connect via native SDKs and REST APIs to warehouses, BI tools, orchestrators, and operational databases; lineage is computed by parsing query history and dbt manifests rather than requiring manual declaration.

The Opinionated Take

Atlan is winning the modern catalog war, and the scoreboard keeps widening. Among companies picking a catalog for the first time in 2024–2026, Atlan wins a disproportionate share of the sexy logos — fintechs, marketplaces, SaaS companies, and anyone who proudly calls their stack "modern data stack." The wins come largely at the expense of Alation (losing the middle) and Collibra (losing tech customers), while DataHub wins the subset of companies that want open-source and have the platform engineers to operate it.

The reasons are not subtle. Atlan ships faster, looks better, deploys in days rather than months, integrates with the tools cloud-native teams actually use, and is sold by a team that understands the buyer is now a Head of Data, not a Chief Data Officer with a governance committee.

Where Atlan still loses. It loses in heavily regulated industries where Collibra's policy workflows, business glossary, and 15-year vendor relationships are non-negotiable. It loses where the buyer demands open-source and DataHub has the stronger story. It occasionally loses on price against Select Star, Secoda, and Castor at smaller accounts. And it loses where the data estate is 90% legacy on-prem systems that Atlan doesn't prioritize (mainframes, SAP BW, Informatica ETL) and the incumbents have deeper connectors.

The long-term risk for Atlan is not a competitor but category convergence: if catalogs, semantic layers, observability, and access control all merge into a single "data intelligence platform," Atlan has to either expand into adjacent categories (which it is doing) or risk being commoditized by warehouse-native alternatives like Databricks Unity Catalog and Snowflake Horizon, both of which are increasingly adequate for customers already committed to one platform.