Introducing DataHub Cloud v2.3

Running an analytics agent is easy. Getting it to answer correctly, every time, is where most teams get stuck. DataHub Cloud v2.3 makes context production-ready as the foundation for accurate and trusted answers from analytics agents. It includes the public beta of DataHub Cloud Context for all customers, deeper evals, a feedback loop that captures what agents get wrong, and metrics fully wired into lineage. Measure context quality from day one and ensure agents provide consistently accurate answers even as your data evolves.

Let’s take a look at what’s new at DataHub.

DataHub Cloud: Context (Public Beta)

The Context module (formerly known as DataHub Context Platform) is now in public beta for every DataHub Cloud customer. It handles the full lifecycle of keeping agent context accurate: generating meaning from query history and BI dashboards, routing proposed changes to domain experts, running evals, and closing the loop on what agents get wrong in production. You can now:

• Generate context from query history and BI dashboards through a guided setup flow
• Trace a failing eval to the exact context document that caused it and fix it in place
• Group evals by domain so each team owns their benchmark questions
• Import results from external eval platforms into the same workflow
• Surface feedback from live agent conversations so production gaps get fixed before they repeat

Learn more about DataHub Cloud: Context.

Context Feedback

When agents get something wrong in production, Context Feedback captures it and routes it to the right person to fix. You can now:

• See conversation feedback alongside scheduled evals in a single Validation view
• Route flagged gaps to the right reviewer without a separate tool
• Fix what agents get wrong before it repeats

Context Feedback is available in Private Beta. Learn more in our docs.

Metrics as first-class lineage citizens (Public Beta)

Metrics and Semantic Models are now searchable and fully wired into lineage. When a metric returns the wrong number, follow the path from chart to metric to physical dataset without leaving the graph. You can now:

• Search for metrics and semantic models alongside tables, dashboards, and datasets
• Trace lineage from a chart through its metric to the physical datasets it reads
• See column-level lineage down to the exact upstream column a metric formula depends on
• Give agents access to validated metric definitions and their lineage as part of the context graph

Learn more about metric lineage in our docs.

Ask DataHub: charts, proposals, and agent-to-agent via MCP

Ask DataHub gets three upgrades in v2.3: inline data visualization, AI-proposed curation changes, and agent-to-agent queries via MCP. You can now:

• See bar charts, line graphs, and metric highlights rendered inline in chat
• Review AI-proposed edits to descriptions, tags and domains rather than authoring from scratch
• Query Ask DataHub or any custom agent via MCP from Snowflake Cortex, Databricks Genie, Claude, Cursor, or any MCP-compatible runtime
• Make agents visible to all users, not only admins

Learn more about Ask DataHub in our docs.

New in User and Platform Experience

• Switch to dark mode across the app (Public Beta)

• Upload your logo, set brand colors, and add a custom favicon

• Use DataHub in Simplified Chinese or Russian (public beta)

New connectors

Two new sources join DataHub’s 150+ connector library. IBM/HCL Informix brings table and schema metadata from Informix instances into the catalog. Monte Carlo ingests observability signals and data quality results alongside your existing catalog metadata, so reliability and discovery live in one place.

What’s new in DataHub Cloud v2.3

Capability	What it does	Who will benefit the most
Context (Public Beta)	Generates context from query history and BI dashboards, routes it for expert review, runs evals, and surfaces feedback from live agent conversations	Data platform engineers, analytics engineers, domain experts
Context Feedback (Private Beta)	Captures missing or incorrect context flagged in live agent conversations and surfaces it for review and resolution	Data platform engineers, domain experts
Metrics as first-class lineage citizens	Metrics and Semantic Models are searchable and fully connected to lineage, with column-level precision from chart to physical dataset	Analytics engineers, data platform teams
Ask DataHub: charts, proposals, and MCP	Renders data as inline charts, proposes curation changes for expert review, and exposes Ask DataHub to other agents via MCP	Data analysts, data platform engineers, domain experts
Dark Mode (Public Beta)	A dark color theme available across the full app	All users
Custom Branding	Drag-and-drop logo upload, accessible brand colors, and a dedicated Favicon field	DataHub admins
Internationalization	Simplified Chinese and Russian now supported in public beta	Global teams

Let’s build together

We’re building DataHub Cloud in close partnership with our customers and community. Your feedback helps shape every release. Thank you for continuing to share it with us.