# Atlan

> Metadata as context for AI, not just a catalog for people

Atlan is a commercial metadata platform covering discovery, lineage, glossaries, contracts, and policies across 80-plus sources, delivered as SaaS with a self-deployed runtime that keeps metadata extraction inside the customer's own network.

Website: https://atlan.com

## Use it when

- You want a complete commercial catalog (discovery, column-level lineage, glossary, contracts, policies) that stands up in weeks rather than quarters.
- Teams program against metadata: SDKs in Python, Java, and TypeScript, plus an MCP server exposing the catalog to AI agents.
- Sources sit behind a firewall but SaaS is acceptable: the Self-Deployed Runtime crawls inside your network and ships results to the tenant.
- Policy should follow tags: a purpose attaches rules to a tag, so one rule covers every column carrying it.

## Think twice when

- SaaS is disqualifying; Atlan is not self-hostable, only its extraction layer is.
- You need list prices; the pricing page is a sales form.
- Running open source is acceptable and someone can operate it; OpenMetadata and DataHub cover the core for free.

## How it runs

Managed SaaS. Connectors crawl 80-plus sources; the optional Self-Deployed Runtime runs as containers on ECS, Cloud Run, or Kubernetes, writes extracted metadata to your own object storage, and transfers it over HTTPS. Quality verdicts from Monte Carlo, Soda, and Anomalo aggregate into trust badges beside Atlan's own contracts.

## Details

### General

| Attribute | Value |
| --- | --- |
| License | Commercial |
| Deployment | Managed |
| Workload | Interactive, Batch |
| Operational complexity | Low |
| Pricing | Subscription |
| Documentation | https://docs.atlan.com |

### Data Catalogs

| Attribute | Value |
| --- | --- |
| Metadata model | A SaaS metadata store you extend rather than model from scratch, assets carrying certificates, tags, terms and custom metadata sets, with badges promoting chosen properties onto the overview. Connectors crawl the sources, and a Self-Deployed Runtime can do that crawling inside your network before shipping results to the tenant |
| Connectors | 80+ data sources by Atlan's own count, warehouses, BI tools and pipeline systems including Fivetran, dbt, Matillion and Airflow through OpenLineage, with a recommended run order of extract-load before transform. The figure is the vendor's; no per-release support matrix exists to check it against |
| Lineage | Column-level and crawled, across Snowflake, Redshift, BigQuery and the rest, with pipeline tools contributing their own, Fivetran processes carry column mappings, and dbt's missing column lineage is reconstructed by parsing the materialised SQL. Impact reports export to CSV, Sheets or Excel |
| Search and discovery | Structured search filtered by source, certificate, owner, tags and glossary term, where the certificate is the trust signal (Verified, Draft, Deprecated or none) and a metadata completeness score says what is still undocumented |
| Business glossary | Glossaries of terms grouped into hierarchical categories, linked to assets, usable as a search filter and importable in bulk, a first-class object rather than a documentation page |
| Policy and compliance | Personas and purposes, which is the distinctive part, a persona curates a set of assets and governs what a team may do with them, while a purpose attaches policy to a tag, so one rule covers every column carrying it rather than one rule per column. Metadata and domain policies sit alongside; the most restrictive rule wins and an explicit deny beats any permit |
| Quality checks | Native and federated at once, data contracts on tables, views and data product output ports, a Data Quality Studio for Databricks, and a trust layer that aggregates verdicts from Monte Carlo, Soda and Anomalo into badges and scores rather than only computing its own |
| Access requests | Governance workflow templates automate requesting, approving and revoking access, and the same machinery covers metadata change requests, which override permissions, so even a user with edit rights goes through the request once change management is on |
| Custom metadata | Custom metadata sets with badges promoting chosen properties onto the asset overview, and SDKs in Python, Java and TypeScript where a badge is simply another asset with the usual CRUD; packages and an app framework extend it further |
| Billing unit | Nothing published: atlan.com/pricing redirects to a sales form, and no licence tiers are documented either, so neither price nor unit can be recorded. The same answer Collibra gives, and less than Alation, which at least publishes its licence levels |

## Capabilities

- [Data Catalogs](https://matca.io/capabilities/data-catalog)

---

Source: https://matca.io/atlan
Last updated: 2026-08-31T11:40:48.062Z