Issues with Default Motherduck MCP Setup Causing Claude Connection Failures

Scott J. · 2026-03-30T23:59:40.135Z

Hi folks - may have this wrong, but - we are having clients set up a default motherduck mcp in their claude desktop instances or claude code - it seems to "pick" a default database and that default seems to stick. When deleted on MD, Claude wigs out and refuses to connect. Is there a client side default being set do you think, or is there anything server-side that assigns a default at mcp setup/runtime?

Issues with Default Motherduck MCP Setup Causing Claude Connection Failures | MotherDuck Community

Tseegi (.
·
·
Hello Scott J.! Thanks for reaching out! A couple of clarifying questions to help us understand what's happening:
1.
What attach mode are you using on the accounts connected to Motherduck MCP?
2.
When you mention deleting on MD, which specific databases are you deleting? Is it my_db, or are there others?
Scott J.
·
·
I don't know which mode, perhaps that's the issue. we didn't delete any my_db databases, but several clients seem to have picked also_analytics on their own (unless we inadvertantly made them do it via claude - maybe alphabetical?)
Tseegi (.
·
·
Hmm.. it looks like my_db was dropped all the way back in January and default database selection is falling back to also_analytics. Could you check that by calling "List all my databases on MotherDuck" and confirm if my_db is available?
Scott J.
·
·
⏺️ The error is: ▎ Catalog Error: Catalog 'also_analytics' has been deleted It seems like a database called also_analytics was previously attached to your MotherDuck session but has since been deleted. This is blocking all queries from executing since MotherDuck tries to load it on connection. You may need to detach or remove the reference to also_analytics in your MotherDuck account. You could try: 1. Going to the MotherDuck web UI and checking if also_analytics appears in your attached databases 2. Or running a DETACH also_analytics command to remove the stale reference Want me to try running a DETACH via the read-write query tool?
Scott J.
·
·
That's what I get back from Claude
Scott J.
·
·
I can't seem to shake free from it and two other people are having the same issue
Scott J.
·
·
I can't run a DETACH because the error comes first
Tseegi (.
·
·
Ok, could you log into your account on MotherDuck UI and confirm if my_db database exists, run USE my_db and create new connection on MotherDuck MCP to see if that makes any difference?
Scott J.
·
·
Scott J.
·
·
no change - I'm using the web mcp, no default database string, I don't understand how it's getting assigned to a dead database
Scott J.
·
·
Is there a way to reset my workspace? it seems like it's attaching in workspace mode but because it can't reconnect to that database, it blocks everything else
Scott J.
·
·
I can attach in single mode
Blossom C.
·
·
Hey Scott J. Tseegi is out today. Would you be able to jump on a quick call to show us what you're seeing? Btw you have a channel on the Motherduck Slack - motherduck-ridealso . This channel has more people monitoring
Scott J.
·
·
Good morning! Sorry I posted on the wrong channel yesterday - just fatigue, we rolled a ton of changes yesterday. Anyway - to get past the error I rehydrated the missing table and the MCPs appear to have started working again. So I think perhaps I just need better guidance on how to gracefully deprecate tables when people have access to them and try to reconnect via the persistent MCP connection
Blossom C.
·
·
Thanks for the update. I'm going continue in the motherduck channel.

Issues with Default Motherduck MCP Setup Causing Claude Connection Failures

20 comments

Issues with Default Motherduck MCP Setup Causing Claude Connection Failures

20 comments