Data catalogs and governance platforms compared (October 2026)
Facts come from the linked vault notes (each with its own sources and dates). “Not verified” means I could not source it on 2026-10-07; it does not mean the feature is absent.
| Platform | Type | Open source | Owner / maker | MCP server | AI governance and lineage-for-AI |
|---|---|---|---|---|---|
| Purview | Catalog plus security/compliance | No | Microsoft | Not verified | DSPM for AI, audit and retention of prompts, labels, DLP; Foundry, Copilot in Fabric, Agent 365 |
| Unity Catalog | Platform-native governance | OSS version, Apache 2.0 (LF AI & Data sandbox) | Databricks | Managed MCP servers, MCP Services as UC securables | Unity Gateway: models, agents, MCP tools as securables; payload logging, budgets; lineage to models and dashboards |
| datahub | Metadata platform | Core Apache 2.0 | DataHub (originated at LinkedIn) | Official, managed and self-hosted, read and write tools | Column lineage, assertions, context for agents; no model registry verified |
| openmetadata | Metadata platform | Apache 2.0 | Collate maintains | Built in, on by default | Lineage, data contracts, quality tests, AI SDK; commercial AI agents via Collate |
| atlan | Active metadata, “context layer” | No | Atlan | MCP Server listed in docs | Context Agents Studio, “AI asset lineage” (docs); registries not detailed |
| collibra | Governance suite | No | Collibra | MCP Server product page on collibra.com (2026-10-07; launch date not shown) | AI Command Center (2025-05-06), ISO 42001 certification and EU AI Act tool (2025-01-22); Maestro, Live Map, Guardian Agents announced 2026-09-23 (Collibra newsroom lists the launch; GA not checked); the 2025 dates are from the vault note and were not seen on collibra.com |
| alation | Catalog, “agentic data intelligence” | No | Alation | Not verified | Agent Studio (2025-10, from the vault note), Analytics Agent; AI Governance, Ontologies, Intelligent Feeds, Console, Governed Collections and Semantic Model Mastering announced as an AIOS expansion on 2026-09-17 (Alation newsroom); GA unchecked |
| informatica | Data management suite (IDMC, CLAIRE) | No | Salesforce since 2025-11-18 | Not verified | Feeds Agentforce 360 and Data 360; named features not checked |
| ataccama | Quality plus catalog plus MDM | No | Ataccama | MCP interface claimed | Data Trust Index, AI steward agent (vendor claims) |
Data quality and observability specialists (not catalogs): monte-carlo-data (closed, ML observability and agent observability), soda-data-quality (Soda Core v4, source-available under the Elastic License 2.0 per the repo LICENSE although the README calls it open source; data contracts), great-expectations (GX Core Apache 2.0, code-first tests).
What AI-ready data needs, and who covers it
- Quality: Monte Carlo, Soda, GX, Ataccama, built-in checks in Purview, Unity Catalog, OpenMetadata, DataHub.
- Lineage (including from data to models and agents): Unity Catalog (data to models, services, dashboards), DataHub and OpenMetadata (column level), Atlan “AI asset lineage”; Purview lineage between data products and assets.
- Access policy: enforced in the data plane by Unity Catalog and the platform; Purview governs access requests and honours existing permissions in Copilot; independent catalogs mostly govern metadata and sync policies (opinion).
- PII handling: Purview classification and labels (opinion: the broadest AI-interaction controls of those listed; not benchmarked), Unity Catalog classification and masks; others via tags (sources show tagging, depth not compared).
- Semantics: glossary and data products (Purview), metric views (Unity Catalog, see databricks-ai-bi), context layers (Atlan, DataHub, OpenMetadata).
Reading across (opinion)
- Platform-native catalogs (Purview, Unity Catalog) win on enforcement where the data lives; independent catalogs win on cross-platform neutrality, and Atlan, Collibra and Alation all list the native catalogs as competitors.
- Informatica joining Salesforce changes its neutrality pitch.
- Open source options with MCP today: DataHub, OpenMetadata, Unity Catalog OSS (MCP support there not verified); Soda Core is source-available (Elastic License 2.0).
Sources
Vault notes linked above, plus checked on 2026-10-07 (primary): raw LICENSE files on GitHub (DataHub, OpenMetadata, Great Expectations, Unity Catalog: Apache License 2.0; Soda Core: Elastic License 2.0); github.com/unitycatalog/unitycatalog README (sandbox project of LF AI & Data); github.com/open-metadata/OpenMetadata README (Collate offers the managed enterprise version and MCP tools); https://www.informatica.com/about-us/news/news-releases/2025/11/20251118-salesforce-completes-acquisition-of-informatica.html and https://www.salesforce.com/news/press-releases/2025/11/18/salesforce-completes-acquisition-of-informatica/ (completion dated 2025-11-18); https://www.collibra.com/company/newsroom and https://www.collibra.com/products/mcp-server; https://www.alation.com/news/; https://www.atlan.com/ (Atlan MCP server mentioned). Other rows rest on the linked vault notes and were not rechecked.
Open items
- MCP servers for Purview, Alation, Informatica (Collibra’s exists, see table); model/agent registry specifics for catalogs other than Unity Catalog; analyst rankings; price comparison deliberately omitted.