Marquez

Marquez is an open-source metadata management service that collects, aggregates, and visualizes data lineage across modern data ecosystems.

Reviewed by 7wData

On this page

Publisher review

Marquez is an open-source metadata management service that collects, aggregates, and visualizes data lineage across modern data ecosystems. Originally created by WeWork in 2017 and open-sourced in December 2019, it is now a graduated project of the LF AI & Data Foundation as of September 2023. The platform serves as the reference implementation of OpenLineage, an open standard for capturing data provenance across pipelines.

Marquez works out-of-the-box with Apache Airflow, Apache Spark, Apache Flink, dbt, and Dagster, tracking dataset provenance, job dependencies, and run execution metrics through a three-component architecture: a metadata repository (graph database), HTTP API, and web UI. It's designed for teams needing lightweight, real-time lineage tracking to understand data origins, debug pipeline failures, and perform impact analysis. The platform is strictly a metadata backend and lineage service, not a full data catalog—it lacks built-in discovery, glossaries, and governance features that other tools provide.

As of 2026, most production deployments using OpenLineage events are routing to Atlan, DataHub, or OpenMetadata instead of Marquez, as those platforms offer richer governance surfaces on the same event stream. Marquez suits organizations with strong internal data platform engineering teams willing to manage operational overhead in exchange for flexibility and cost savings.

Get the AI & data signal, daily.

335k+ subscribers read this every morning. One email, both newsletters. Unsubscribe anytime.

How it works

  1. Real-time lineage API

    OpenLineage-compatible HTTP endpoint that collects metadata from running jobs and applications in real-time, enabling immediate visibility into data flow.

  2. Visual graph interface

    Web UI that renders complex data dependencies as interactive graphs, allowing users to trace dataset lineage, explore job runs, and examine performance metrics.

  3. RESTful and GraphQL metadata queries

    Both REST API (production-ready) and GraphQL endpoint (beta) enable programmatic access to metadata for automation, backfill orchestration, and root-cause analysis.

  4. Modular three-tier architecture

    Decoupled metadata repository (PostgreSQL-backed graph database), API layer, and UI; extensible design allows custom integrations and third-party tool plugging.

  5. Multi-tool integration

    Works natively with Apache Airflow, Spark, Flink, dbt, and Dagster via OpenLineage; Python and Java client libraries provided for custom integrations.

  6. Dataset, job, and run tracking

    Maintains provenance of how datasets are consumed and produced, logs job execution records with runtime metrics, and preserves immutable run history for compliance.

  7. Scalable graph database backend

    Built on PostgreSQL with graph-aware schema; designed to handle millions of entities and scale horizontally in Kubernetes via Helm charts.

Strengths and trade-offs

Strengths

  • Pure open-source (Apache 2.0); no vendor lock-in and full control over deployment.
  • Lightweight and purpose-built for lineage only; lower resource footprint than all-in-one catalogs.
  • Deep OpenLineage integration as the reference implementation; real-time event capture with low latency.

Trade-offs

  • Lineage-only tool; does not include data discovery, business glossaries, data quality, or governance—requires pairing with other tools.
  • Operational burden: requires Docker, Java 17, PostgreSQL 14+, and Kubernetes expertise; no managed SaaS option available.
  • Limited user-facing features; UI is functional but not designed for business stakeholders; API-first approach demands engineering resources.

Pricing context

Marquez is free and open-source under the Apache License 2.0. There is no commercial licensing, managed SaaS offering, or support tier. Organizations self-host and operate the platform, bearing all infrastructure, deployment, and maintenance costs.

Community support is available via the LF AI & Data Foundation mailing list, GitHub issues, and Slack channel. Enterprise support is not officially offered by the Marquez project.

Alternatives

User reviews

No user reviews yet. Be the first to write one.

Sources

Reporting on this tool draws on these publicly available sources.

  1. marquezproject.ai — Official Marquez project site; features, architecture, deployment options, and integrations
  2. github.com — GitHub repository; open-source code, license (Apache 2.0), active releases, current version (0.50.0 as of October 2024)
  3. atlan.com — Comprehensive 2026 analysis of Marquez architecture, features, use cases, founding by WeWork in 2017, open-sourced 2019, strengths and weaknesses, comparison with alternatives
  4. lfaidata.foundation — LF AI & Data Foundation project status; graduated September 2023, initially contributed by WeWork December 2019, governance structure, community resources
  5. atlan.com — Market positioning of Marquez in 2026; comparison with OpenMetadata, DataHub, and Atlan; production deployment trends
  6. dev.to — Comparative analysis of Marquez vs DataHub, OpenMetadata, and other lineage tools; feature and architecture differences