Status, September 2026: Amundsen is archived. The Amundsen repository carries the notice “Due to inactivity, this project was archived in September 2026. The contents will remain available for historical purposes.” GitHub records the archive on September 10, 2026. The Apache-2.0 code stays public and forkable, but the repo is read-only: no new releases, no security patches, no maintainer review of issues or pull requests. This comparison used to turn on a dormant roadmap. It no longer does. DataHub is still shipping; Amundsen has stopped. Everything below about how Amundsen works remains accurate as of its last release, and none of it is a reason to start a new deployment. OpenMetadata is the other maintained open-source option worth putting next to DataHub.
What is the difference in the architecture of Amundsen and DataHub?
Amundsen vs. DataHub: An overview
DataHub and Amundsen are tools developed as internal projects at LinkedIn and Lyft to address their organization’s data cataloging and discovery use cases.
DataHub was open-sourced in 2020, and the original authors founded a company, AcrylData, that maintains the project.
Amundsen, on the other hand, was open-sourced in 2019 and donated to LF AI & Data for incubation. Stemma, the company that offered a managed version of Amundsen, was acquired by Teradata in 2023.
Both DataHub and Amundsen have a microservice-based architecture, in which different services handle the frontend, the backend, metadata ingestion, full-text search, and other components.
Amundsen vs. DataHub: Architecture
DataHub has a more complex architecture, partly because it is designed to scale to support enterprise use cases. It has four external dependencies:
Amundsen, on the other hand, has a simpler architecture and fewer external dependencies with:
- Neo4j/Apache Atlas
- Elasticsearch
This makes Amundsen deployments easier, as compared to DataHub.
Key takeaway: Amundsen is easier to use and deploy, lacks certain features, and is archived, so that feature gap is now permanent. DataHub is slightly more complex to manage and is enterprise-ready and scales well with a larger user base.
What are the key features and focus areas of DataHub and Amundsen?
Both DataHub and Amundsen were created as internal projects to solve very similar problems; hence, there’s a significant overlap in the feature set between the two tools.
While they support the same set of core features, Amundsen and DataHub exhibit different levels of maturity.
Some features, such as data asset popularity and user statistics, are available out of the box in Amundsen. The same features are available in DataHub, but may require additional setup. The same is true for other features that work natively in DataHub but not in Amundsen.
One difference outranks the rest: DataHub is under active development and Amundsen is archived. Every Amundsen cell in the table below describes the project as of its final release. Every DataHub cell describes something that can still change. Now, let’s look at the key features and how they compare.
What's the difference? | Amundsen (archived September 2026) | DataHub (active) |
|---|---|---|
Architecture | Uses either Neo4j or Apache Atlas as the metadata storage backend, and Elasticsearch for text-based discovery. | Uses MySQL as the primary database, Kafka for communication between services (besides REST APIs), and Elasticsearch for full-text search. Has more external dependencies than Amundsen. |
Search and discovery | Simpler, but limited, search interface. Has less contextual search and discovery capability than DataHub. | Many UX-enhancing features for search and discovery are available, but some of them require configuration. |
Data assets supported | Mainly supports relational database, data warehouse, or data lake-type data assets in schemas, tables, and files. It also supports dashboards, but in a very limited way | Supports all types of data assets, including schemas, tables, data pipelines, dashboards, AI/ML models. Also supports data products as a concept. |
Ingestion methods | With a Neo4j backend, Amundsen only supports batch ingestion workflows using Airflow as the orchestrator and Databuilder as the underlying connector library. Streaming ingestion works only when you're using Apache Atlas as the metadata storage engine. | Supports more than 50 connectors for metadata ingestion. Uses Kafka for streaming ingestion. |
Data quality | Can integrate with external data quality tools, but only to a very limited extent. | Can integrate with third-party data quality tools like dbt and GX. |
Data lineage | Supports table-level lineage natively. It can also support column-level lineage, but requires fairly complex configuration and setup. | Offers column-level lineage natively, especially by extracting lineage information from tools like dbt and Airflow. |
Data governance | Offers very basic support for tagging and descriptions, which doesn't help create a better data governance experience than DataHub. | Data is secured using multiple lines of defence using sensitive data tagging, access policies, fine-grained access control, and data domains. |
What are the deployment and maintenance considerations for DataHub and Amundsen?
While DataHub is a more complex tool to deploy, it offers more streamlined, thoroughly tested deployment patterns for both on-premises and cloud-based environments. Here’s a quick comparison of these options for both tools:
Deployment Method | DataHub (active) | Amundsen (archived September 2026) |
|---|---|---|
Local or On-Premises | Docker, k8s/Helm | Docker, k8s/Helm |
AWS | EKS (official) | ECS (uses Docker) |
Azure | AKS (official) | No official support |
GCP | GKE (official) | No official support |
Amundsen has limited deployment options. DataHub, on the other hand, supports all mainstream deployment options for local development and testing and for production environments, both on-premises and in the cloud.
All of DataHub’s deployment options are well-maintained and receive official support. Amundsen’s official deployment options were tested, are not up to date, and will not be updated again. Azure, GCP, and AWS EKS were community territory even before the archive. With the repository read-only, that community has nowhere upstream to land a fix, so treat those paths as unsupported rather than lightly supported.
Key takeaway: DataHub is harder to deploy but production-ready across clouds. Amundsen is simpler upfront, and since the archive nobody is maintaining the deployment guides you would be following.
How mature are both DataHub and Amundsen, and what does their future roadmap look like?
There are stark differences between the maturity and roadmaps of the two tools. The following table captures those differences as they stand on the 21st of September 2026.
Future roadmap criteria | DataHub | Amundsen (archived September 2026) |
|---|---|---|
Project status | Active | Archived 10 September 2026; repo is read-only (did not graduate) |
Maintainer | None. Issues and pull requests are closed to review (formerly Stemma) | |
Documentation | Frozen (docs still state "Vision for 2021") | |
Latest release date | 14 August 2024, and the last one |
Amundsen has no future roadmap, and the September 2026 archive makes that permanent. DataHub’s release schedule and roadmap are frequently updated. Pre-release notes for the next release candidate, v1.4.0, are already out.
On maturity, development, and maintenance activity, Amundsen is out for new production use. Teams that have run it for years and know the codebase can keep it running. What they are keeping running is a fork in all but name, and the security patching that used to arrive upstream is now on them.
Know your current data catalog maturity so you can pick the right platform
Take assessmentHow should you choose between Amundsen and DataHub?
Based on the evidence, DataHub is the choice for a data cataloging and discovery tool. Amundsen is simple and easy to manage, and it is archived, so there is no version of a new deployment that ends well.
Stay on Amundsen only if:
- You already run it internally and have maintained custom extensions, and you accept that maintaining the project itself is now your job.
- Your needs are lightweight: basic search, simple metadata documentation, minimal governance, and no expectation of new features.
- You have a plan for security patches, because none are coming from upstream.
If none of those hold, this is a migration question rather than a comparison. DataHub and OpenMetadata are the maintained open-source destinations.
Amundsen vs DataHub decision tree. Source: Atlan.
Choose DataHub if:
- You are solving for enterprise maturity (scalability, reliability, roadmap).
- You require column-level lineage, streaming metadata ingestion, etc.
- You need multiple deployment options
- Enterprise governance features (tags, domains, ownership, etc.)
Key takeaway: Continue using Amundsen only if your organization already has the expertise to build and maintain the project, which since the archive means maintaining the project itself. Otherwise, pick DataHub.
A more realistic decision: It’s rarely just Amundsen vs. DataHub
In practice, the decision you’ll likely face isn’t DataHub vs. Amundsen, but rather a comparison of all the key candidates that address data cataloging and discovery for your organization.
Both Amundsen and DataHub built solid foundations for search and discovery. Only one of them is still being built on.
But neither delivers advanced governance, cross-system automation, or AI-ready context at the depth modern enterprises often require.
This is where broader platforms like Atlan enter the evaluation. They unify metadata, lineage, governance, and AI context into a single experience that supports enterprise-scale trust, compliance, and cross-team adoption.
Autodesk chose Atlan to activate their data mesh with Snowflake, after having worked with Amundsen.
“When we got to our data mesh initiative in 2021, we decided to select Amundsen. Some of the drawbacks though, being open-source, were a lot of gaps in functionality. It turned out to be a lot of work adding basic features that we needed, like the ability to update metadata by a data owner. We had to build our own UI to do that, or to add things like lineage. If we wanted to do that with Amundsen, it was an investment.” - Mark Kidwell, Chief Data Architect, Data Platforms and Services
With Atlan, Autodesk’s data teams got the primary starting point to find the data they need and immediately start using it, with capabilities, such as:
- An out-of-the-box setup with Snowflake (Autodesk’s data lake)
- Custom metadata related to data quality and ownership
- Open API access to integrate their vast data ecosystem
- Strong UX driving broader adoption – technical and business users
Ready to choose the right metadata platform for your team?
Choosing the right tool depends on your scale, roadmap, and governance maturity. Between these two, new deployments should go to DataHub. Amundsen is archived.
Amundsen can still serve lightweight use cases for teams already deeply invested in maintaining it. Those teams now own the codebase in practice, and a project that cannot receive a security patch upstream is hard to justify for a growing data estate.
And if your organization needs capabilities that go beyond open-source catalogs—such as active metadata, advanced governance, or AI-ready context—explore platforms like Atlan in a broader evaluation.
FAQs about Amundsen vs. DataHub
1. Which tool should you choose, Amundsen or DataHub?
For new deployments, DataHub, because Amundsen was archived in September 2026. The Amundsen repository is read-only: no new releases, no security patches, no maintainer review of issues or pull requests. DataHub is actively developed and enterprise-ready, with streaming ingestion, native column-level lineage, and fine-grained governance.
Amundsen still handles lightweight search-and-discovery for teams that already run it, and the Apache-2.0 code stays available. From now on those teams carry the maintenance and the security work themselves.
Many teams also evaluate commercial platforms such as Atlan when they require active metadata, AI-powered search, and advanced governance beyond what open-source tools provide.
2. What is Amundsen and how does it function as a data discovery tool?
Amundsen is an open-source data discovery tool developed by Lyft. It helps organizations manage their metadata by providing a user-friendly interface for searching and discovering data assets. Amundsen uses an ETL framework for metadata ingestion, so teams can catalog and access their data. The project was archived in September 2026 and its repository is now read-only, so the tool works as described but receives no further development.
3. How does Amundsen compare to DataHub in terms of features and usability?
Amundsen is quick to stand up and simple to use, which is what made it attractive to small teams. Since the September 2026 archive that simplicity comes with a frozen feature set and no upstream support.
DataHub, developed by LinkedIn, offers more advanced governance features, supports a wider range of integrations, and is still shipping releases.
4. What are the main differences between Amundsen and DataHub regarding integration capabilities?
Amundsen supports a variety of data sources and has a straightforward integration process, though its connector set is frozen at the September 2026 archive and will not gain new sources.
DataHub offers more extensive integration options, including support for GraphQL and Kafka, and keeps adding connectors, which suits organizations with complex data ecosystems.
5. How do Amundsen and DataHub support data lineage tracking?
Both Amundsen and DataHub provide data lineage tracking features.
Amundsen allows users to visualize data lineage through its catalog, while DataHub offers advanced lineage capabilities, including column-level lineage tracking, which helps organizations understand data flow and transformations.
6. When should you choose a commercial tool over Amundsen or DataHub?
You should consider a commercial metadata platform when your needs extend beyond basic discovery into enterprise governance, automation, and AI readiness. This includes requirements like:
- Active metadata management
- Automated, actionable, cross-system column-level lineage at scale
- Automated policy enforcement and tag propagation
- Faster time-to-value
- Broad adoption across business users.
Commercial platforms (like Atlan) also reduce operational burden by providing managed deployments, support, and a clear product roadmap. Open-source tools typically require significant in-house engineering effort to match that, and with Amundsen archived since September 2026, its share of that effort now includes maintaining the codebase itself.
