Claude DataHub connectorWhat Claude can do in your DataHub account.
The Claude DataHub connector exposes 34 tools. 14 read your catalog, 19 change something in it, 1 is documented nowhere. Below: what Claude finds, what it can annotate, the switch without which no write exists, and who can shut it all off.
Verified Trustpilot reviews · AI, automation & growth agency
What it changes for a data team
Answering "where does this number come from?" used to mean opening DataHub, finding the table, expanding lineage and reading through queries. Now you ask Claude: it searches the catalog, walks the lineage and shows you the real queries that hit the table. If writes are switched on, it can also tag, document and grow the glossary.
Size up a change before making it. get_lineage lists what depends on a column, and get_lineage_paths_between shows every transformation between two assets.
Write SQL that matches real usage. get_dataset_queries pulls the queries teams actually run, then draft_sql_for_tables proposes a query grounded in those joins and filters. Both only query DataHub; neither changes the catalog.
Keep governance current. add_tags, add_terms and add_owners classify and assign in bulk across several entities, columns included, while propose_create_glossary_term sends a term to review instead of forcing it in.
What Claude will not do: it runs no SQL, it only drafts it. The 19 write tools only exist once the modification option is enabled on the DataHub side. And nothing fires when a new table appears; scheduled processing is a job for an automation tool, a different line of work gathered on the Integrations hub.
The vocabulary in one minute
Five words worth knowing before you connect DataHub.
- Connector
- The link you set up once between Claude and an account you already have, so Claude can work in it while it answers you.
- Tool
- One named action a connector opens to Claude. Claude picks which ones to call; the directory lists them by name.
- Authorization
- The service's own sign-in screen, where you hand Claude the access it will use. Granted once per person, and revocable.
- Approval
- The confirmation Claude waits for before an action that changes your account, shown in the chat right when it matters.
- MCP
- The shared standard behind connectors: it is what lets an assistant like Claude talk to an outside service in the first place.
Connect DataHub to Claude in three steps
- 01
Find DataHub in Claude
In Claude's settings, open Customize, then Connectors, and look for DataHub. On a Team or Enterprise workspace, an Owner or Primary Owner enables it for the organization first, then each person connects.
- 02
Sign in to your instance
Click Connect. DataHub's window asks for your domain, then your login, SSO included if your tenant uses it. If the link breaks later, Disconnect and connect again.
- 03
Review the authorization
Read the authorization screen before you accept: it belongs to DataHub, not Claude, and it sets what the access covers. You can withdraw that access later from your DataHub account.
The 34 tools, grouped by what they do
DataHub gives Claude 34 tools: 14 that read your account, 19 that change something in it, and 1 no official source describes.
Three groups: what Claude reads, what it changes, and what no source describes. Names stay exactly as Claude displays them.
- 14 read
- 19 write
- 1 not documented
What Claude reads (14)
14 toolsSearch, lineage, queries and glossary history, with nothing modified.
compare_glossary_term_versions
Puts two versions of a glossary term side by side and shows what changed between them. DataHub documents it alongside get_glossary_term_versions, which lists the full history.
draft_sql_for_tables
Drafts a SQL query against the tables you name, grounded in what DataHub knows about them: schemas, sample queries and lineage. The query is a proposal to review before you run it.
get_dataset_queries
Brings back real SQL queries that hit a dataset or a column, whether hand-written or system-generated. You see the joins, filters and aggregations people actually use.
get_entities
Retrieves detailed metadata for one or more entities from their URN, in batches when needed. It is what follows a search when Claude has to inspect several results at once.
get_glossary_term_versions
Shows the version history of a glossary term, so you can see when and how its definition evolved from one revision to the next.
get_lineage
Walks lineage upstream or downstream for any entity, table, column or dashboard, with filters, pagination and hop control. Claude sees where data comes from and what depends on it.
get_lineage_paths_between
Traces the exact path between two assets or two columns, intermediate transformations and SQL included. Where overall lineage gives you a map, this gives you the route.
get_me
Returns details about the signed-in user: profile and group memberships. Like every read tool here, it only queries DataHub and never changes the catalog.
grep_documents
Searches inside document content with regular expressions, to find a specific piece of information spread across several notes.
list_lifecycle_stages
Lists the lifecycle stages configured for glossary terms, such as proposed, approved or deprecated. Per DataHub, it only queries the catalog and never modifies it.
list_pending_proposals
Shows the metadata proposals still waiting for a reviewer. You get the queue of what is pending a decision. Deciding on them goes through accept_or_reject_proposals, a different tool.
list_schema_fields
Goes through a dataset's schema fields with keyword filtering and pagination. Useful when search truncates the column list or a schema is very wide.
search
Queries DataHub with structured keyword search, boolean logic, filters, pagination and optional sorting by usage. Its results can then be inspected in batches with get_entities.
search_documents
Finds documents by keyword, filterable by platform, domain, tag, glossary term or owner. It targets the catalog's notes and guides, not its tables.
What Claude changes (19)
19 toolsTags, terms, owners, domains, descriptions, glossary and proposals. No source gives them a confirmation of their own, so the general rule applies.
accept_or_reject_proposals
Approval: see the ruleAccepts or rejects pending metadata change proposals. It is the decision that closes a review cycle.
add_owners
Approval: see the ruleAssigns owners to entities, with the ownership type you want, such as technical owner or data owner.
add_related_terms
Approval: see the ruleLinks glossary terms together as synonyms, containment or inheritance, to express relationships in the business vocabulary.
add_structured_properties
Approval: see the ruleSets structured properties, typed metadata fields, on entities. DataHub accepts string, number, URN, date and rich text values.
add_tags
Approval: see the ruleAdds tags to entities or to specific columns, in bulk across several entities if needed.
add_terms
Approval: see the ruleAttaches glossary terms to entities or columns, tying them to a business definition or a data classification.
create_glossary_term
Approval: see the ruleCreates a new glossary term directly, without going through a proposal.
create_glossary_term_version
Approval: see the ruleRecords a new version of an existing term to keep track of how it changes over time.
propose_create_glossary_term
Approval: see the ruleSubmits a new glossary term as a proposal awaiting approval, instead of adding it straight away.
propose_lifecycle_stage
Approval: see the ruleProposes a change to an entity's lifecycle stage, subject to approval, rather than applying it immediately.
remove_domains
Approval: see the ruleRemoves an entity from its domain. DataHub notes that an entity belongs to a single domain at a time.
remove_owners
Approval: see the ruleRemoves ownership assignments from entities, whatever their type.
remove_structured_properties
Approval: see the ruleDeletes structured properties set on entities, once a typed metadata field is no longer needed.
remove_tags
Approval: see the ruleStrips tags from entities or columns, in bulk if needed.
remove_terms
Approval: see the ruleDetaches glossary terms from entities or columns when the link was wrong or no longer makes sense.
save_document
Approval: see the ruleSaves standalone documents to DataHub's knowledge base: insights, decisions, FAQs, notes. They are filed under a configurable parent folder.
set_domains
Approval: see the ruleAssigns an entity to a domain. Each entity belongs to one domain only.
set_lifecycle_stage
Approval: see the ruleSets the lifecycle stage of an entity or glossary term directly, for example proposed, approved or deprecated.
update_description
Approval: see the ruleUpdates, appends to or removes the description of an entity or a column, with markdown formatting.
Undocumented (1)
1 toolThe directory publishes this name and nothing more. No official text says what it does, and this page will not guess.
note_metadata_observation
This name appears in the directory listing and nowhere else: DataHub's documentation, which covers every other tool on the sheet, does not describe it. This page will not say more than that.
What Claude asks before it acts
By default, Claude stops and asks for your go-ahead before each action it takes on an account for you. That rule is what covers the 19 write tools.
DataHub annotates every write tool so compatible clients can flag that it changes the catalog and ask for confirmation, but no source describes tool-specific behavior. On a Team or Enterprise workspace, owners decide whether a member may let certain actions through without a prompt, and can keep reads open while shutting writes for the whole organization. Each person also signs in with their own DataHub login, so Claude sees nothing beyond what they can.
Which plans it is available on
None of the 819 sheets in the directory shows plan availability. That information is not published anywhere, connector by connector.
The general rule is published: remote connectors are open to all users on Claude, Cowork, Claude Desktop and mobile. On Team and Enterprise, an Owner or Primary Owner turns the connector on for the organization first. On the DataHub side, what matters is the version of your instance, detailed under limits. For the current state on your account, the sheet in the official directory remains the reference.
Where this connector stops
A connector is not an automation. Claude calls these tools while it answers you, and nothing starts on its own when a table changes.
Only the publisher documents this connector: neither Claude's help center nor Anthropic's developer docs have a page on it. The tool list is an observed floor, not a promise, and the partner badge is not a security audit. Anthropic says so on every sheet: it does not choose the tools a publisher exposes and does not guarantee they behave as described. Connect only what comes from a publisher you know.
Need help connecting DataHub to Claude?
A person reads every message.
