OpenDataDiscovery Platform

OpenDataDiscovery Platform (ODD) is an open-source catalog for finding and governing data products across data teams. Users can search and filter metadata by attributes such as source, owner, and tag, then explore ownership and links to related sources. ODD tracks lineage across data objects and microservices, imports data-quality test results, and displays dataset quality statuses. It also catalogs machine-learning experiment metadata and lineage, and supports logging experiments to connect data science and engineering work. Listed integrations include MongoDB, ClickHouse, dbt, Snowflake, Tableau, BigQuery, Kafka, and PostgreSQL. Teams can deploy locally with Docker Compose, use Kubernetes and Helm charts, or choose AWS Marketplace QuickLaunch. A user interface and REST API provide access to collected metadata. ODD supports LDAP authentication and USER and ADMIN roles; its documentation warns against using simple login-form authentication in production. Attachments to metadata entities are limited to 20 MB, while links can point to larger or remotely stored files. The platform is free and open source, with public development and listed team and community support options.

ODD combines catalog search with lineage, data-quality, and machine-learning metadata features, plus several deployment options. Teams should account for its attachment size limit and follow its guidance on production authentication.

ODD is intended for data teams of any size, including enterprises, small and midsize companies, and data science teams. It may suit groups seeking catalog search, lineage, quality status, and machine-learning metadata in one platform.

OpenDataDiscovery Platform (ODD) is an open-source catalog for data assets, services, and machine-learning metadata. It fits data teams of any size that need shared governance across engineering and data science, provided they are prepared to run and secure the platform themselves.

Overview

ODD combines catalog discovery with end-to-end data-object lineage and microservices lineage. Search and faceted filters can narrow assets by source, owner, or tag, while ownership information and related-source links help people understand who manages an asset and what it connects to. This is more useful than a simple dataset index when teams must trace dependencies across pipelines and services.

The platform imports data-quality test results and displays dataset quality status, so existing checks can be visible alongside catalog context. It also treats machine-learning entities as first-class discovery objects: collectors can catalog experiment metadata and lineage, and users can log experiments. That shared view is well suited to collaboration between data science and data engineering.

ODD uses an open metadata collection standard to connect catalogs, tools, and feature stores. Lookup Tables support creating and maintaining reference data for governance. After the first ETL execution, owners of affected entities receive alerts, giving teams a trigger for follow-up when upstream changes affect their assets.

Key features

Discovery and lineage

Faceted search makes large catalogs more navigable, while tags, ownership, and related links provide practical context for handoffs and investigations. Pipeline lineage, report lineage, and impact analysis extend the catalog from “what exists” to “what could be affected” by a change. End-to-end lineage for both data objects and microservices is particularly relevant to organizations whose products span data processing and service layers.

Quality and machine learning

Imported quality-test results let teams expose dataset health in the same place as metadata. The benefit depends on feeding ODD those results; it is not a replacement for a testing system. Experiment metadata, experiment logging, and ML lineage give data scientists and engineers a common record of work instead of isolating machine-learning context from the wider catalog.

Integrations, interface, and governance

Listed integrations include MongoDB, ClickHouse, dbt, Snowflake, Tableau, BigQuery, Kafka, and PostgreSQL, among others. A web user interface is paired with a REST API for collected metadata, supporting both interactive discovery and programmatic workflows. LDAP authentication is supported, and the access model defines USER and ADMIN roles, with ADMIN granted all permissions by default. Production deployments should avoid the simple login-form authentication that the documentation warns against.

Deployment and storage

Teams can run ODD locally with Docker Compose, deploy it on Kubernetes through Helm charts, or use the AWS Marketplace QuickLaunch. Those choices accommodate different operating models, but they also make deployment, upgrades, and security part of the buyer’s responsibility. Attachments on metadata entities are limited to 20 MB; larger or remotely stored files should be represented with links instead.

Pricing

The Open-source platform plan costs free. It is free and open source, with public development described as forever open-source. A free plan and free trial are available, and no seat limits, usage quotas, renewal charges, or trial duration are stated. The absence of a license fee suits teams that can supply their own infrastructure and administration; it does not remove those operational costs.

Platforms

ODD is offered through an API and web interface and can run on Linux or as a self-hosted deployment. Docker Compose, Kubernetes Helm charts, and AWS Marketplace QuickLaunch cover local, cluster-based, and AWS-oriented installation paths.

Who it's for

Enterprises can use the lineage, role model, integrations, and federated metadata approach to coordinate broad data estates. Small and midsize companies may value the zero-price open-source model when they already have technical staff to operate a catalog. Data science teams benefit from ML experiment metadata and lineage alongside conventional data assets. ODD is a weaker fit for organizations seeking a fully managed catalog with minimal deployment and authentication work.

Pros and cons

Pros

• Broad asset coverage: Data objects, microservices, pipelines, reports, quality results, and ML experiments can be understood in one catalog.
• Flexible deployment: Docker Compose, Helm-based Kubernetes, and AWS QuickLaunch support different infrastructure preferences.
• Open integration model: An open metadata standard and numerous listed connectors help connect catalogs, tools, and feature stores.
• No software fee: The free, open-source plan removes licensing charges for teams able to operate the service.

Cons

• Operational burden: Self-hosting means deployment, maintenance, and production security remain with the team.
• Authentication caution: The simple login form is not recommended for production, so teams need a stronger LDAP-based setup.
• Attachment cap: Metadata attachments stop at 20 MB, requiring links for larger or remote files.
• Quality is ingestion-dependent: Dataset status reflects imported test results rather than replacing the systems that generate them.

Alternatives

For a broader comparison of lineage-focused products, see Data Lineage Tools.

Coalesce Catalog is the better choice when a freemium catalog with defined hosted plans is preferable: its Developer plan is free for one user with 2,000 capped monthly actions, one project, and one environment, while Starter costs $150 a month.

Databricks Notebooks suits teams centered on notebook-based work that want a free serverless workspace, with paid usage available without upfront costs.

Grai is an alternative for teams wanting a fully open-source, self-hosted product with control over data and hosting.

Amazon DataZone fits organizations that prefer a paid, AWS-oriented service billed according to requests, metadata storage, compute, and AI recommendation tokens.

Atlas is a freemium, open-core option for teams whose priority is database migration tooling rather than a broad discovery catalog.

Collate offers a freemium lineage and catalog route with self-hosted and web platforms.

Google Cloud Knowledge Catalog is intended for teams choosing a paid Google Cloud service with usage-based processing charges.

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

Verdict

Choose OpenDataDiscovery Platform if your data organization needs one open-source home for catalog search, lineage, quality signals, and ML metadata, and can own deployment and security. Its main advantage is breadth without a license fee; look elsewhere if you want a managed experience or cannot accommodate production authentication work and the 20 MB attachment ceiling.