OpenDataDiscovery Platform

OpenDataDiscovery Platform (ODD) is an open-source catalog for data and machine-learning assets. It is best for teams that need shared discovery across data engineering and data science; its breadth is appealing, but self-hosting and production authentication require care.

Overview

ODD combines searchable metadata with end-to-end data-object and microservices lineage, imported data-quality results, and machine-learning experiment metadata. Teams can filter by source, owner, and tag, then follow ownership details and links to related sources. That makes it more useful for tracing relationships across data products and services than a catalog focused on finding datasets alone.

An open metadata collection standard connects catalogs, tools, and feature stores. Lookup Tables let teams maintain reference data for governance, while owners of affected entities receive alerts after the first ETL execution. The alert timing is useful to know when planning change follow-up. Attachments are capped at 20 MB; links provide a practical alternative for larger or remotely stored files.

Key features

Lineage, quality, and ML metadata

ODD supports pipeline and report lineage alongside impact analysis, helping teams understand dependencies rather than merely browse asset descriptions. It can import test results and show dataset quality statuses, so its quality view depends on results being supplied by the team's testing workflows. For data-science collaboration, ML entities are treated as part of discovery: the platform catalogs experiment metadata and lineage generated by collectors and supports experiment logging.

Deployment and access

Teams can deploy with local Docker Compose, Kubernetes Helm charts, or AWS Marketplace QuickLaunch, and use the interface or REST API to work with collected metadata. That range suits organizations choosing between local evaluation and managed infrastructure they operate themselves. LDAP authentication is supported, but the documentation warns against using simple login-form authentication in production. USER and ADMIN roles provide a basic division of access, with ADMIN holding all permissions by default; teams needing finer-grained permissions should assess that model against their requirements.

The integration set includes MongoDB, ClickHouse, dbt, Snowflake, Tableau, BigQuery, Kafka, and PostgreSQL. The project describes itself as forever open-source with public development. Support options include a dedicated team, a helpful team, and a community.

Pricing

The Open-source platform plan is free. It is free and open-source, making it a fit for teams that can host and operate the platform without a subscription charge. Automated discovery, impact analysis, pipeline lineage, and report lineage are supported. No seat or usage cap is stated for this plan, so the decision turns less on a paid-tier comparison than on whether the team's deployment and security responsibilities are acceptable.

A free trial is also offered, though no trial length or separate paid plan terms are given. Buyers should treat ODD as a self-hosted open-source choice rather than assume a time-limited commercial tier.

Platforms

ODD supports API, Linux, self-hosted, and web use. Docker Compose, Helm-based Kubernetes deployment, and AWS Marketplace QuickLaunch give teams several documented deployment paths; the REST API complements the user interface for metadata access.

Who it's for

ODD suits data teams of any size, from small and midsize companies to enterprises, that want discovery spanning data engineering and data science. It is particularly relevant where lineage, quality-test visibility, and ML experiment context belong in the same catalog. Teams that do not want to deploy infrastructure, need production authentication without configuration decisions, or routinely attach files over 20 MB should weigh those constraints before adopting it.

Pros and cons

Pros: Search, lineage, quality results, and ML metadata share one catalog, supporting investigation across data assets and experiments.

Pros: Free, open-source deployment options include Docker Compose, Kubernetes, and AWS QuickLaunch, giving teams control over how they run it.

Pros: The REST API and open metadata collection standard connect the catalog to tools, catalogs, and feature stores.

Cons: Production use requires deliberate authentication choices; simple login-form authentication is discouraged in the documentation.

Cons: Attachments stop at 20 MB, so larger files need to live elsewhere and be linked.

Cons: Dataset quality visibility relies on importing test results, and owners' alerts follow the first ETL execution rather than being described as immediate.

Alternatives

For a wider comparison, see Data Lineage Tools.

Atlas is worth considering if its open-core model, Apache 2.0 core, and advanced-feature split better match your requirements.

Coalesce Catalog may suit teams that prefer a freemium catalog with a one-user Developer plan capped at 2,000 monthly actions, one project, and one environment, or a $150 a month Starter plan.

Grai is an alternative for teams looking specifically for a fully open-source, self-hosted project that keeps data and hosting under their control.

Databricks Notebooks may fit teams centered on notebooks that want a free edition with one serverless workspace and limited compute size and usage.

Egeria is another free option.

Apache Atlas is a free metadata-governance alternative with no paid plans or usage limits stated.

Collate may appeal to teams considering OpenMetadata's open-source core alongside a UI and connectors under a Community License.

Alex Solutions is an alternative for buyers seeking an annual enterprise platform with unlimited users, sources, features, and integrations; it has no free plan.

Verdict

Choose ODD if your team wants a no-subscription, open-source catalog that brings data lineage, imported quality results, and ML metadata together. Its strongest case is shared discovery across engineering and data science; look elsewhere if you need hands-off hosting, more nuanced default access control, or attachments above 20 MB.