--- title: "DataHub Pricing: What Open Source and Cloud Cost to Run" url: "https://atlan.com/know/ai-agent/context-layer/datahub-pricing/" description: "DataHub pricing: Core is free under Apache License 2.0 and Cloud is priced by quote. What self-hosting runs, what patches cost in time, and a sizing worksheet." author: "Emily Winks" author_role: "Data Governance Expert" published: "2026-11-05" updated: "2026-11-05" --- ## How much does DataHub cost? DataHub Core is free to deploy under Apache License 2.0. DataHub Cloud is scoped to your use case and data environment, and DataHub asks you to contact sales for pricing. Its pricing depends on data volume, users, and the capabilities you need, so the real comparison is the cost of running Core yourself, including its servers, dependencies and upgrade time, against a Cloud quote. Four cost lines decide the total: - **Core license:** free under Apache License 2.0 - **Cloud quote:** scoped to data volume, users, and capabilities - **Self-hosting work:** 4 components, 4 dependencies, ordered upgrades and patches - **Agent context:** the definitions and policy rules agents read at runtime ## Key takeaways - DataHub Core is free under Apache License 2.0; DataHub Cloud is scoped by quote to data volume, users, and capabilities. - DataHub counts 4 components and 4 external dependencies to self-host, plus release-train upgrades and patches. - Price the context your AI agents read as its own line next to the catalog, whichever DataHub option you choose. --- DataHub Core, the open source metadata project, is free under Apache License 2.0, and DataHub prices Cloud by quote. DataHub's deployment guide counts 4 main components and 4 external dependencies for a self-hosted Core. Atlan's Context Layer for AI covers the cost line that sits beside either option, the governed context your agents read, through a four-stage Context Pipeline. ### Build Your Own Shortlist Matrix Give it your candidate tools, your stack, and why you're evaluating. It returns a matrix scored on connector coverage, where lineage stops, who maintains the metadata, and what happens if you leave. [Read the skill](/skills/catalog-tool-shortlist.md). *Paste into a new chat* ``` Use the skill at https://atlan.com/skills/catalog-tool-shortlist.md to evaluate a data catalog tool shortlist against my stack. Ask me for whatever it needs. ``` *Run once in a terminal* ``` curl -fsSL --create-dirs \ -o ~/.agents/skills/catalog-tool-shortlist/SKILL.md \ https://atlan.com/skills/catalog-tool-shortlist.md ``` *For an agent* ``` curl -fsSL https://atlan.com/skills/catalog-tool-shortlist.md ``` Human Agent ClaudeChatGPTGeminiCursorOther Copy A free license and a quote-based product put the cost in different places. With Core, the cost is the servers, the dependencies and the engineers who keep them current. With Cloud, it is a number you learn on a sales call, so your own sizing is what you bring to that call. Either option also needs someone to fund the context your AI agents read at runtime, whichever tool holds it. The [enterprise context layer](https://atlan.com/know/what-is-the-enterprise-context-layer/) is where Atlan keeps that line, and the [build vs buy vs bundle TCO](https://atlan.com/know/ai-agent/context-layer/context-layer-tco-build-vs-buy-vs-bundle/) breakdown prices it next to the platform. - Start with the license and the Cloud quote: [Is DataHub free to use?](#is-datahub-free-to-use) - Then the work of running Core: [What does it take to self-host DataHub?](#what-does-it-take-to-self-host-datahub) - Then the worksheet for the quote call: [How do you size DataHub cost before asking for a quote?](#how-do-you-size-datahub-cost-before-asking-for-a-quote) | What it is | Value | |---|---| | License of Core | Apache License 2.0 | | Price of Core | Free (open source), per DataHub's Core vs Cloud page | | Price of Cloud | Scoped to your use case and data environment; contact sales | | What drives the Cloud quote | Data volume, users, and the capabilities you need | | What self-hosting runs | 4 main components and 4 external dependencies, per DataHub's Kubernetes guide | | Patch on the v1.6.0 LTS line | v1.6.0.4, released 2026-10-29 | --- ## Is DataHub free to use? DataHub Core is free to download and deploy under Apache License 2.0, and DataHub Cloud is a separate paid product. The [LICENSE file on DataHub's master branch](https://github.com/datahub-project/datahub/blob/master/LICENSE) opens with "Apache License Version 2.0, January 2004". [DataHub's Core vs Cloud page](https://datahub.com/products/cloud-vs-core/) says it directly: "DataHub Core is open source and free to deploy." Under the [Apache License text](https://www.apache.org/licenses/LICENSE-2.0), each contributor grants a no-charge, royalty-free copyright license. The license is free. The servers, the dependencies and the people who keep both running are still yours. Where those costs land comes next. --- ## How much does DataHub Cloud cost? DataHub scopes DataHub Cloud pricing to your use case and data environment and routes the price question to sales. The cost row on [DataHub's Core vs Cloud page](https://datahub.com/products/cloud-vs-core/) reads "Scoped to your use case(s) and data environment. Contact sales for pricing." Its FAQ names the drivers: data volume, users, and which capabilities you need. [DataHub's Getting Started with DataHub Cloud docs](https://docs.datahub.com/docs/managed-datahub/welcome-acryl) show Cloud instance URLs on the acryl.io domain. DataHub also runs an [ROI estimator](https://datahub.com/roi/) that projects benefits from your inputs. It estimates value, not price, so the quote stays the only price in the picture. Run the quote with a [data catalog RFP template and evaluation criteria](https://atlan.com/know/data-catalog/data-catalog-rfp-template-with-evaluation-criteria/) in hand, and settle early whether you are buying a catalog or a context layer, as [data catalog vs context layer](https://atlan.com/know/data-catalog-vs-context-layer/) lays out. --- ## What does it take to self-host DataHub? Self-hosting DataHub means running its 4 components and 4 external dependencies yourself, as [DataHub's Kubernetes deployment guide](https://docs.datahub.com/docs/deploy/kubernetes) lists them. The guide names GMS, an optional MAE Consumer, an optional MCE Consumer and the Frontend as the main components. It says the dependencies must be deployed before DataHub itself, and that they can run on-prem or come as managed services. The graph index can be Neo4j or Elasticsearch, so how many separate systems you operate depends on what you choose and what your platform team already runs. | Layer | What DataHub's docs list | Who operates it when self-hosted | |---|---|---| | DataHub components | GMS, MAE Consumer (optional), MCE Consumer (optional), Frontend | Your team | | Message queue | Kafka | Your team, or a managed service you buy | | Primary database | MySQL, Postgres, or MariaDB | Your team, or a managed service you buy | | Search index | Elasticsearch; v1.8.0 notes add OpenSearch 3.x, supported next to OpenSearch 2.x and Elasticsearch 8.x | Your team, or a managed service you buy | | Graph index | Neo4j, or Elasticsearch | Your team, or a managed service you buy | | Ingestion | A Python library, run with `datahub ingest` | Your team | ### The four components [DataHub's components overview](https://docs.datahub.com/docs/components) describes the Metadata Store as a Spring Java service backed by MySQL, Elasticsearch and Kafka for primary storage and indexing. Ingestion sits outside that service. It is a Python library you run with `datahub ingest`. Mapping these layers onto your own estate is easier with the [context catalog](https://atlan.com/know/context-catalog/) framing in hand. ### The four dependencies and who runs them Each dependency is a line on your bill or your on-call rota. If your platform team already runs a Kafka cluster, a database or a search cluster, the cost to count is the capacity and operations you add to it, and a dedicated cluster is a different line from a shared one. DataHub's [architecture overview](https://docs.datahub.com/docs/architecture/architecture) says the system is proven to 10M+ assets and O(1B) relationships at LinkedIn and other companies in production; that is DataHub's own statement, and a load test with your own asset counts is the check. If agents will read this stack over MCP, the [MCP server implementation guide](https://atlan.com/know/mcp-server-implementation-guide/) belongs on the same reading list. Before you size anything, list the catalog workloads you run today, starting from [data catalog examples](https://atlan.com/know/data-catalog/data-catalog-examples/). --- ## What do upgrades and security patches add to the cost of self-hosting? A self-hosting team inherits DataHub's release train: ordered upgrade steps, runtime version requirements and security patches, all stated in [DataHub's release notes](https://docs.datahub.com/docs/releases). Four entries show the shape of the work: - **v1.6.0.4**, released 2026-10-29, is a patch on the v1.6.0 LTS line. The notes say its focus is security, with October dependency bumps for known CVEs. - **v1.6.0.3**, released 2026-09-25, requires Python 3.11 on that hotfix line for the ingestion CLI, images and related packages, and drops Python 3.10 for v1.6.0.x. - **v1.7.0.1**, released 2026-09-03, sets an order. From a version before 1.7.0, you upgrade to v1.6.0 first with Helm chart 1.0.3 and let the system-update job finish. Then you move to v1.7.0.1 with Helm chart 1.1.0, and the notes say "Do not skip 1.6.0." - **v1.8.0** adds OpenSearch 3.x, supported next to OpenSearch 2.x and Elasticsearch 8.x, which makes the search index a version decision as well as a hosting one. Each entry is staff time to plan, test and run in sequence, and the ordered path is the line to budget for. If you ingest lineage through OpenLineage, add that integration to the regression list; [DataHub OpenLineage support](https://atlan.com/know/ai-agent/data-for-ai/datahub-openlineage-support/) covers the pairing. --- ## DataHub Core vs DataHub Cloud: what changes in cost and operations DataHub's own comparison puts the same open metadata model under both options, and moves installation, scaling, upgrades and patching from your team to DataHub when you choose Cloud. | Line | Core | Cloud | |---|---|---| | Cost | Free (open source) | Scoped to your use case(s) and data environment; contact sales | | Operational overhead | Manual install, configuration, and upgrades; your team owns uptime and troubleshooting | Guided implementation and onboarding; managed by DataHub engineers | | Support | Community Slack and self-service docs | Dedicated customer success team and a private Slack channel | | Uptime commitment | Owned by your team | 99.5% uptime SLA | | Moving between them | Core to Cloud: yes, per DataHub | Runs on the same open metadata model as Core | | Source connectors | 140+ | 140+ | | Lineage | Column-level lineage | Column-level lineage | | MCP | GraphQL and MCP retrieval | GraphQL and MCP retrieval, plus hosted and scoped MCP servers | | Context documents | Listed in both | Listed in both, plus context documents generation | | Agent features | Skills library | Skills library, plus the Ask DataHub AI agent | Rows come from [DataHub's Core vs Cloud page](https://datahub.com/products/cloud-vs-core/) and its [comparison guide](https://docs.datahub.com/docs/managed-datahub/managed-datahub-overview). For the next step, [DataHub alternatives](https://atlan.com/know/ai-agent/context-layer/datahub-alternatives/) and the [open source vs commercial data catalog total cost of ownership](https://atlan.com/know/data-catalog/open-source-vs-commercial-data-catalog-total-cost-of-ownership/) work through the same trade. --- ## How do you size DataHub cost before asking for a quote? Sizing DataHub cost comes down to six lines, filled from your own bills and time sheets, with the Cloud line left for the quote. | Cost line | What to count | Where the number comes from | |---|---|---| | Infrastructure for the 4 components | Cluster capacity for GMS, the Frontend and any consumers you run, in each environment you stand up | Your cloud bill | | The 4 dependencies | Kafka, a database, a search index and a graph index, self-run or as managed services, plus backups and monitoring for each | Your cloud bill or managed-service quotes | | Upgrade and patch time | Engineer hours per release step in DataHub's notes, including reindexing and testing | Your team's time sheet from the last upgrade | | Cloud quote | Data volume, users and capabilities | DataHub's quote | | Connectors in scope | Your sources counted against DataHub's list of 140+ | Your source inventory | | Agent context | Who writes, reviews and refreshes the definitions agents read | Your AI program plan | The totals are two formulas with no figures in them. Core is infrastructure plus dependencies plus upgrade hours times your loaded hourly rate plus agent context. Cloud is the quote plus the connectors and network work on your side plus agent context. Take a hypothetical estate of 20 sources, a count you replace with your own. List each source against DataHub's connector list before the call, then count the `datahub ingest` runs you will schedule, one per source. The same lines work as a [build vs buy evaluation](https://atlan.com/know/ai-agent/context-layer/metadata-tooling-build-vs-buy-evaluation-criteria/). The agent-context line is the one most worksheets leave blank, and [how much context is enough](https://atlan.com/know/how-much-context-is-enough/) is the place to start filling it. --- ## What to ask on a DataHub quote call Because DataHub scopes Cloud pricing to data volume, users and capabilities, five questions make the quote comparable. | Ask | Why it matters | Tied to | |---|---|---| | What unit does the quote scale on? | Volume, users and capabilities move the price differently | DataHub's FAQ on cost drivers | | Which Cloud-only capabilities sit inside the quote? | The comparison guide splits features between Core and Cloud | DataHub's comparison guide | | Are the hosted and scoped MCP servers included? | Agents reach the catalog through them | DataHub's comparison guide | | What does the uptime commitment cover? | DataHub states a 99.5% SLA | DataHub's Core vs Cloud page | | Is the in-VPC remote ingestion agent included? | It decides what runs inside your own network | DataHub's comparison guide | Put the answers next to the worksheet. A [context engineering platforms comparison](https://atlan.com/know/context-engineering-platforms-comparison/) gives you a second set of criteria for the evaluate step, and [context graph vs knowledge graph](https://atlan.com/know/context-graph-vs-knowledge-graph/) helps when the vendor's answers lean on graph vocabulary. --- ## Where does the cost of context for AI agents sit? The cost your agents care about is context, meaning definitions, lineage and policy rules served at the moment they act, and it sits as its own line next to catalog cost under either DataHub option. DataHub's comparison guide lists hosted MCP servers, scoped MCP servers and context document generation as Cloud features, and lists GraphQL and MCP retrieval for both options. Those rows tell you how an agent can read the catalog. The people cost sits elsewhere: someone writes and reviews the definitions, and someone answers when an agent quotes a stale one. Put an owner and hours against that line before any agent goes to production, whichever catalog you run. The reasons are laid out in [why AI agents need an enterprise context layer](https://atlan.com/know/why-ai-agents-need-an-enterprise-context-layer/), and the practice is [context engineering](https://atlan.com/know/what-is-context-engineering/). A [semantic layer](https://atlan.com/know/semantic-layer/) holds metric definitions, and a [context graph](https://atlan.com/know/what-is-a-context-graph/) connects them to lineage and policy. Running agents at volume has its own bill, covered in the [cost to run AI agents at scale](https://atlan.com/know/ai-agent/cost-to-run-ai-agents-at-scale/). --- ## How Atlan approaches the cost of context for agents Atlan treats context as the line you price, and builds it as a Context Layer for AI that agents read at runtime. The worksheet above covers the platform and the people. The quote call settles the Cloud line, and the agent-context line needs an owner either way. Atlan answers the agent-context line with the four-stage Context Pipeline: Unify, Bootstrap, Collaborate, Activate. Unify connects your warehouses, BI tools and business systems. Context Agents draft definitions, metrics and ontology from what is already there, and the [context bootstrapping](https://atlan.com/know/context-bootstrapping/) approach gives people a draft to review instead of a blank page. In Context Engineering Studio, people approve what becomes the canonical definition. The Context Lakehouse then serves certified context to agents through MCP, SQL and APIs, and [what Atlan MCP is](https://atlan.com/know/what-is-atlan-mcp/) covers the MCP route. The result is one Enterprise Data Graph of definitions, lineage and policy context that agents read. Teams with a catalog already in place can [sit a data catalog on top of existing catalogs](https://atlan.com/know/data-catalog/can-a-data-catalog-sit-on-top-of-existing-catalogs-without-replacing-them/), and the [how to implement an enterprise context layer](https://atlan.com/know/how-to-implement-enterprise-context-layer-for-ai/) guide walks the rollout. For the wider field, see the [agentic data catalog](https://atlan.com/know/agentic-data-catalog/) and [enterprise AI context management tools](https://atlan.com/know/enterprise-ai-context-management-tools/) comparisons. --- ## Which worksheet line is still blank? Two facts are settled: DataHub Core is free under Apache License 2.0, and DataHub Cloud is priced by quote against data volume, users and capabilities. Between them sit the infrastructure, the dependencies and the upgrade time you can fill from your own bills, and the quote you fill on the call. The agent-context line has to be priced either way, and the [RAG pipeline vs context layer TCO](https://atlan.com/know/ai-agent/rag-pipeline-vs-context-layer-tco/) comparison shows one way to do it. Which of your six lines is still blank? --- ## FAQs about DataHub pricing ### 1. Is DataHub free to use? DataHub Core is free to download and deploy under Apache License 2.0. DataHub Cloud is a separate paid product that DataHub scopes to your use case and data environment. A free license still leaves you the servers, the dependencies, and the staff time to run it. ### 2. How much does DataHub cost? DataHub Core costs nothing to license. For DataHub Cloud, DataHub says pricing depends on data volume, users, and the capabilities you need, and it directs buyers to contact sales for a quote. Your own total for Core is the infrastructure and engineer time you spend to run it. ### 3. Is DataHub open source? Yes. DataHub Core is published under Apache License 2.0, and the license file sits in the project's public repository. DataHub Cloud is a separate managed product that runs on the same open metadata model as Core, according to DataHub. ### 4. What does DataHub Cloud add over Core? DataHub says Cloud handles installation, onboarding, scaling, adoption support, upgrades, and security patching, with a 99.5% uptime SLA and a dedicated customer success team. Its comparison guide also lists hosted and scoped MCP servers and the Ask DataHub agent as Cloud features. ### 5. Is DataHub reliable? DataHub's architecture page says the system is proven to 10M+ assets and O(1B) relationships at LinkedIn and other companies in production, and DataHub Cloud carries a 99.5% uptime SLA. Both are DataHub's own statements, not an independent reliability test, so run your own load test before you rely on them. ### 6. Can you move from DataHub Core to DataHub Cloud without losing your setup? DataHub says yes. Its comparison page states that Cloud runs on the same open metadata model as Core, so your existing ingestion, lineage, and glossary work carries over rather than starting from scratch. Confirm the migration steps and timeline with DataHub before you commit. --- ## Sources 1. [DataHub Core vs. DataHub Cloud: Open Source vs Enterprise, DataHub](https://datahub.com/products/cloud-vs-core/) 2. [LICENSE (Apache License Version 2.0), DataHub project on GitHub](https://github.com/datahub-project/datahub/blob/master/LICENSE) 3. [Apache License, Version 2.0, Apache Software Foundation](https://www.apache.org/licenses/LICENSE-2.0) 4. [Deploying with Kubernetes, DataHub docs](https://docs.datahub.com/docs/deploy/kubernetes) 5. [DataHub Components Overview, DataHub docs](https://docs.datahub.com/docs/components) 6. [DataHub Architecture Overview, DataHub docs](https://docs.datahub.com/docs/architecture/architecture) 7. [DataHub Releases, DataHub docs](https://docs.datahub.com/docs/releases) 8. [Core vs Cloud: Comparison Guide, DataHub docs](https://docs.datahub.com/docs/managed-datahub/managed-datahub-overview) 9. [Getting Started with DataHub Cloud, DataHub docs](https://docs.datahub.com/docs/managed-datahub/welcome-acryl) 10. [DataHub ROI estimator, DataHub](https://datahub.com/roi/) ## Next steps - [See how Atlan's context layer works in recorded demos](https://atlan.com/context-layer-demo/) - [Book a demo or talk to Atlan's sales team](https://atlan.com/forms/talk-to-sales-contact/)