Claude MotherDuck connectorWhat Claude can do in your MotherDuck account.
The Claude MotherDuck connector exposes 15 tools. 10 read your data warehouse, 5 write to it, 0 undocumented. Claude explores your databases, runs SQL and builds Dives, MotherDuck's interactive visualizations. Here's what it reads, what it changes, and where it stops.
Verified Trustpilot reviews · AI, automation & growth agency
What changes for your data work
MotherDuck is a cloud data warehouse built on DuckDB. With this connector you ask a question in plain English and Claude finds the tables, writes the SQL (the query language databases speak) and can save the result as a Dive, an interactive visualization it hands back as a link.
Find your way around the catalog. When you don't know the exact table names, search_catalog runs a fuzzy search across databases, schemas, tables and columns, then list_columns spells out the structure of the one that matters.
Answer a numbers question. query runs read-only SQL: any operation that would modify data or schema gets rejected. When a table really needs reshaping, query_rw handles the statements that change things.
Ship a visualization. Claude loads the rules with get_dive_guide, saves a Dive with save_dive and hands back a clickable link. share_dive_data then opens the underlying data to your organization.
What Claude won't do: it never starts anything on its own, every query begins in your conversation. Query results get truncated past a cap MotherDuck sets. And the directory sheet doesn't include the other tool families the publisher documents. A scheduled data load is a job for an automation tool, a different line of work, listed on the Integrations hub.
Five words before you start
Quick definitions so the tool list makes sense.
- Connector
- The link you set up one time between Claude and an account you already have, so Claude can use it while it answers you.
- Tool
- A single named capability a connector gives Claude. Claude picks which ones to call, and the directory sheet lists each by name.
- Authorization
- MotherDuck's own sign-in screen, where you give Claude the access it will use afterwards. Granted once per person, and revocable.
- Approval
- The confirmation Claude waits for before it completes an action that changes your account, shown in the chat right when it matters.
- MCP
- The open standard connectors are built on. It's what lets an assistant like Claude talk to an outside service such as MotherDuck.
Set up the Claude MotherDuck connector in three steps
- 01
Find MotherDuck
In Claude's settings, open Customize then Connectors and locate MotherDuck. On a Team or Enterprise workspace, an Owner or Primary Owner first has to enable the connector so members can sign in individually.
- 02
Start the connection
Click Connect on its row, then log in to MotherDuck in the window the service opens itself. If the link breaks one day, use Disconnect and connect again.
- 03
Review the authorization
Read MotherDuck's screen before you accept. The service, not Claude, decides what the access covers. You can also take that access back later from your MotherDuck account itself.
All 15 MotherDuck tools, sorted by what they do
MotherDuck gives Claude 15 tools: 10 that read your account, 5 that change something in it.
Two groups: what Claude reads and what it changes. Names stay exactly as Claude shows them.
- 10 read
- 5 write
What Claude reads (10)
10 toolsTen tools that explore your catalog, run read-only queries or open your Dives.
list_databases
Shows every database your MotherDuck account can reach, both the ones you own and attached shares, with name and type.
list_shares
Displays the database shares other users have sent you, each with a name and a URL. That URL is what lets Claude attach the share as a database through the query tool.
list_tables
Enumerates the tables and views in one MotherDuck database, along with any comments people left on them. Nothing is changed: it only shows what the database holds.
list_columns
Details each column of a table or view, with its type and any comment. Claude grasps the structure before it drafts a query.
search_catalog
Searches the whole catalog (databases, schemas, tables, columns and shares) using fuzzy matching. MotherDuck presents it as the way to discover available data when you don't know the exact names.
query
Executes SQL against your databases without changing anything: MotherDuck rejects any operation that would alter data, schema or account settings.
ask_docs_question
Asks a question about DuckDB or MotherDuck and gets an answer drawn from their official documentation. The answer comes from those docs, not from your databases, which this tool doesn't touch.
get_dive_guide
Loads MotherDuck's instructions for building a Dive. The publisher says to call it before any Dive is saved, so the code that follows matches what MotherDuck expects. It returns guidance only and leaves your account unchanged.
read_dive
Reads one Dive by its ID, including its full JSX/React component code. An optional version number lets it pull a specific earlier version. The Dive itself is left untouched.
list_dives
Lists the Dives you own, from most to least trusted status (Endorsed, Ready, Draft, then Archived), with the most recent first. Archived Dives are excluded unless you ask for them.
What Claude changes (5)
5 toolsFive tools that modify data, Dives or their sharing. No source details their confirmation, so the general rule below applies.
query_rw
Approval: see the ruleRuns SQL statements that can modify data or schema in your databases, the very operations the read-only tool turns away.
save_dive
Approval: see the ruleCreates a new Dive in your workspace and returns a link to open it. Before saving, the tool validates the code for common issues such as invalid SQL or missing exports.
update_dive
Approval: see the ruleChanges the title, description or code of an existing Dive, then returns a link to the updated Dive.
delete_dive
Approval: see the ruleRemoves a Dive from your workspace using its ID. MotherDuck states the action is permanent and the Dive can't be recovered afterwards.
share_dive_data
Approval: see the ruleCreates organization-wide shares for databases you own that a Dive uses, so others in the organization can view it.
What Claude checks with you first
By default, Claude stops and asks before every action it takes on an account for you. Here it covers five write tools.
No source says whether Claude asks before query_rw, save_dive, update_dive, delete_dive or share_dive_data, so the default rule covers them. On Team and Enterprise, workspace owners decide whether members may let some actions through without being asked again, and can keep reads open while closing writes. MotherDuck also documents a way to limit the assistant to read-only access. And Claude works with your MotherDuck rights, nothing more.
Which plans include it
None of the 819 sheets in the official directory shows plan availability. Connector by connector, that information isn't published anywhere.
The general rule is public: remote connectors are open to all users on Claude, Cowork, Claude Desktop and mobile. On Team and Enterprise, an Owner or Primary Owner enables the connector for the organization. For MotherDuck's current status, its sheet in the official directory is the reference to check.
Where the MotherDuck connector stops
A connector isn't an automation. Claude calls these tools while it answers you, and nothing fires when a table fills up with new rows.
Only the publisher documents this connector: neither Claude's help center nor Anthropic's docs have a page for it. The directory badge isn't a security audit, and Anthropic doesn't guarantee a publisher's tools behave as described, so connect only publishers you know. By default its sheet exposes five write tools, and the publisher documents a way to restrict the assistant to read-only access. MotherDuck's docs also describe other tools that aren't on the sheet: this page leaves them out.
Need help connecting MotherDuck to Claude?
A person reads every message.
