Claude Dashboards Connects to Your Data Warehouse — What a Business Decides First

What shipped

Claude Dashboards lets someone ask a question in plain language — “show quarterly revenue by region” — and Claude builds and maintains a live dashboard from it. The feature connects directly to Amazon Redshift, BigQuery, ClickHouse, Databricks, and Snowflake, plus CRM tools like Salesforce. It shipped in beta on paid plans alongside Claude Motion (animated explainers built as editable code, in beta on Team and Enterprise), and alongside Claude Docs, Slides, and Design graduating out of beta onto every plan, including Free, with enterprise additions like Customer-Managed Encryption Key support. The headline capability is straightforward: Claude is no longer limited to generating a document from what you paste in — it can query the system of record and keep the answer current as the underlying data changes.

Why this is a different class of connector

Most connectors a business turns on first are read paths into a shared drive, a calendar, or a project tracker — useful, but bounded to documents one person already had access to. A warehouse connector is different in kind. Snowflake, BigQuery, Databricks, Redshift, and ClickHouse are where a company keeps the joined, modeled version of everything: revenue by customer, usage by account, margin by product line, often assembled specifically so a query can cross tables that were never meant to sit next to each other in a single export. Pointing Claude at that layer is not the same decision as pointing it at a shared folder. It is closer to handing someone a seat in the data warehouse itself — which is exactly why the question of who gets that connector, and under what credentials, deserves more scrutiny than the first wave of connector rollouts usually got.

The transparency feature that makes it governable

To Anthropic's credit, the feature was built with that scrutiny in mind. Every chart on a Claude dashboard shows the query behind it — a viewer can click a number and see exactly how it was computed — and each dashboard displays when its data was last refreshed. That turns a dashboard from a number someone has to trust into a number someone can check. For a finance or ops team deciding whether to let a generated dashboard stand in for a hand-built report, that query-level visibility is the difference between a tool they can put in front of a board and one they can only use for a gut-check. It is also, not coincidentally, the same instinct behind Anthropic publishing its security and compliance posture at the Trust Center rather than asking enterprises to take data handling on faith.

The enterprise controls shipped alongside it

The same release that introduced Dashboards also moved Docs, Slides, and Design out of beta onto every plan, including Free, and added Customer-Managed Encryption Key support for enterprise organizations — letting a company hold the key that protects its own Claude-generated content rather than relying solely on Anthropic-managed encryption. That detail is easy to skip past in a feature announcement, but it is the kind of control a security team actually asks for before a warehouse-connected feature gets anywhere near production data: not just who can query the numbers, but who holds the key to the artifacts Claude produces from them. Anthropic's enterprise terms and the controls available on Claude for Enterprise are the artifact a security review cites; a feature announcement is not.

The decision a business actually has to make

Turning on a warehouse connector is one setting, but it compresses three decisions that usually get made separately:

None of this is a reason to avoid the feature. A live, source-cited dashboard that refreshes itself is a genuine step up from a quarterly export someone forgets to update. It is a reason to treat the warehouse connector the way a mature company already treats BI tool access — scoped, logged, and reviewed — rather than the way a browser extension gets installed.

The broader connector pattern

Dashboards is the latest entry in a fast-growing Claude connectors catalog, and like the rest of that catalog it runs on the Model Context Protocol — the open standard Anthropic introduced and now co-stewards, documented at modelcontextprotocol.io. That matters for the same reason it has mattered with every connector before it: a business standardizing on MCP is standardizing on the protocol the ecosystem converged on, not locking itself into one vendor's proprietary integration format. What changes with each new connector category — calendars, then document stores, now warehouses — is not the plumbing underneath. It is how much is at stake if the scoping decision above gets made carelessly.

Where Skoor fits

Skoor implements and manages Claude on business devices, and connector governance is a standing part of that work: role-based permission profiles with connector allowlisting, so a warehouse connector gets scoped to the team and task that actually need it, and an audit trail of what was connected, by whom, and under what profile. The catalog of what we currently cover is public at Skoor Solutions.