Amundsen vs Apache Atlas.
Amundsen and Apache Atlas both anchor in catalog & discovery — 3 dimensions differ, 6 hold. Below: posture, coverage diff, and capability matrix.
The Lyft-born OSS catalog that invented search-first discovery — historically important, but development has largely stalled since 2024.
The ASF's Hadoop-native metadata framework — typed entities, classification propagation via lineage, and Ranger-enforced policies.
Teams that already run Amundsen and need to understand what they have, or teams with a genuinely minimal requirement — usage-ranked table search and ownership tracking, nothing more — who are comfortable owning a codebase that is no longer moving.
Organisations running Hadoop-era estates — Hive, HBase, Kafka, Impala — especially on Cloudera Data Platform, where Atlas ships as the embedded governance layer.
What each is betting on.
Honest read: maintenance mode. Created at Lyft, open-sourced October 2019, joined LF AI & Data as an incubation project in August 2020. Development has slowed to near-zero — the last release was databuilder 7.5.1 (August 2024) and the last commit to the monorepo (April 2025) moved a maintainer to emeritus status. Stemma, the managed-Amundsen company founded by an Amundsen co-creator, was acquired by Teradata in 2023 and discontinued as a standalone product. The repo is not archived, but treat this as software that is no longer evolving.
Born inside Hortonworks' Data Governance Initiative (announced December 2014, with enterprise partners), entered the Apache Incubator in May 2015 and graduated to a top-level ASF project in June 2017. Today it survives primarily as the governance layer embedded in Cloudera Data Platform, and as the foundation of Microsoft Purview's Data Map, which is based on Atlas and supports Atlas APIs. Release cadence has slowed markedly: 2.3.0 (Dec 2022), 2.4.0 (Jan 2025), 2.5.0 (Apr 2026).
Each tool's current strategic narrative, verbatim from its profile.
How each tool describes the other.
Against datahub and openmetadata, the comparison is mostly historical: both cover Amundsen's discovery core, then add business glossary, governance workflows, column-level lineage, and data quality — with active communities shipping monthly. Amundsen's remaining edge is smallness: if your entire requirement is usage-ranked table search with owners and column stats, Amundsen does that with less conceptual overhead than either successor. Against apache-atlas, Amundsen is the friendlier UX but the less-maintained codebase; Atlas persists inside Hadoop-legacy estates Amundsen never targeted. Against select-star or secoda, the trade is self-hosted OSS versus paying for a maintained SaaS that does the same search-first job.
Against datahub, Atlas is the prior generation of the same idea — typed entity model, Kafka-borne metadata events — with the stronger security-enforcement story and the far weaker modern-stack coverage. Against openmetadata, the connection is literal: OpenMetadata's founding team includes Atlas veterans, and the project is in many ways Atlas rebuilt for the cloud-warehouse era, with 120+ connectors and a simpler stack. Against amundsen, the contrast is purpose — Amundsen is discovery-first and lightweight, Atlas is governance-first and heavy. Teams leaving Hadoop generally migrate to DataHub or OpenMetadata rather than extending Atlas.
Each quote is pulled from the named tool's own "Where it fits" write-up.
Spec sheet diff.
| Amundsen | Apache Atlas | |
|---|---|---|
| Vendor | LF AI & Data Foundation | Apache Software Foundation |
| dbt integration | Plugin | None |
| OpenLineage | Consumer | None |
| Founded | 2019 | 2015 |
Full Amundsen pricing → Full Apache Atlas pricing →
Both share Primary cluster: Catalog & discovery · Deployment: Self-hosted only · License: Open source · Pricing: OSS · paid tiers · Free tier: Yes · OSS self-host: Yes · Status: ● active
Each tool's center of gravity.
| Cluster | Amundsen | Apache Atlas |
|---|---|---|
| Lineage & metadata | 1/3 | 2/3 |
| Quality & testing | 0/3 | 0/3 |
| Catalog & discovery | 3/3primary | 3/3primary |
Scored 0–3 per cluster on the same rubric across all tools. A 0 means the cluster isn't the tool's focus, not that the feature is absent. See the methodology.
Where they cover different ground.
The declared feature set.
2 of 3 declared features differ — listed first.
These are each tool's self-declared key_features; a blank dot means
undeclared, not impossible.
| Feature | Amundsen | Apache Atlas |
|---|---|---|
| Business Glossary Catalog & discovery | ||
| Column-Level Lineage Lineage & metadata | ||
| Table-Level Lineage Lineage & metadata |
Where they disagree.
Catalog & discovery
2 of 9 differ| Amundsen | Apache Atlas | |
|---|---|---|
| Business glossary | ||
| Tag propagation |
Lineage & metadata
3 of 7 differ| Amundsen | Apache Atlas | |
|---|---|---|
| Column-level | ||
| Cross-system | ||
| BI lineage |
When to pick each.
Teams that already run Amundsen and need to understand what they have, or teams with a genuinely minimal requirement — usage-ranked table search and ownership tracking, nothing more — who are comfortable owning a codebase that is no longer moving. The core idea still holds up: index tables, dashboards, and people into Elasticsearch, rank results by query-log usage so the tables analysts actually trust float to the top, and keep the UX focused on the single question "which table should I use?" Databuilder's pull model is plain Python, so extending it from an existing Airflow deployment is straightforward for a data platform team.
Organisations running Hadoop-era estates — Hive, HBase, Kafka, Impala — especially on Cloudera Data Platform, where Atlas ships as the embedded governance layer. The pairing with Apache Ranger is the reason to choose it: classify a column PII in Atlas and Ranger enforces masking and access policies on the actual data, with classifications propagating automatically through lineage as data moves. That metadata-driven-security loop is still the strongest in open source. Also relevant for teams building against Microsoft Purview, whose Data Map is based on Atlas and supports Atlas APIs — the type system and REST surface transfer directly. Vendor-neutral ASF governance and a genuinely complete Apache-2.0 codebase (no open-core split) round out the case.
What each does best.
Amundsen stands out for
- Pioneered usage-ranked, search-first discovery — PageRank-style ranking from query logs remains a genuinely good idea that successors copied
- Deliberately small surface area — analysts get table search, owners, column stats, and previews without a governance platform's learning curve
- Apache-2.0 under neutral LF AI & Data governance, with nothing held back for a paid tier
- Databuilder is plain-Python ETL — custom extractors are easy to write and schedule from an existing Airflow deployment
Apache Atlas stands out for
- Classification propagation via lineage plus Apache Ranger integration — tag-based access control and data masking enforced on the data itself, not just documented in the catalog
- Rigorous typed metadata model (types, entities, first-class relationships) that handles technical and business metadata in one graph
- Vendor-neutral ASF governance and a complete Apache-2.0 codebase — no managed tier holding features back
- Column-level lineage for Hive and Impala, captured in real time by hooks since the 0.8 line — mature and battle-tested within its ecosystem
Tools both also compete with.
All Amundsen alternatives, scored →All Apache Atlas alternatives, scored →
A note on this comparison.
Every capability value above traces to Amundsen or Apache Atlas's own structured spec, which links back to its source — nothing here is averaged or smoothed across the two.
Notice something inaccurate? Send a correction.