Setting Up Supabase MCP for Clay AI Prompts: Need Guidance

·

Hi everyone, does anyone have experience with setting up a Supabase MCP to be compatible with Clay AI prompts? I looked into it and it seems like it would need to be configured as custom MCP server that needs to be hosted on Vercel / Google etc? Is there no out-of-the box that I can connect my Clay with Supabase without having the load the whole database as context into the prompt?

6 comments

Abdul S.
·
·
There isn’t an out-of-the-box Supabase clay setup. you’ll need a custom MCP server hosted on vercel or google easier way is to expose only the data you need through an API and pull it into Clay tables.
1
Alexander F.
·
·
Okay thanks - in your experience is this reliable and works quite well?
Abdul S.
·
·
Yes, it’s reliable if you only pull the fields you need through the API - keeps it fast and stable
Suhaib S.
·
·
Alexander F. no native Supabase MCP in Clay. Either host your own MCP (Vercel/GCP) or just hit Supabase REST/PostgREST API in Clay to pull only needed rows
Alexander F.
·
·
Okay thanks! Unfortunately we would need to get a lot of rows as context so it would be hard / expensive as pure context to the LLM
Suhaib S.
·
·
in that case better to pre-filter in Supabase with SQL, then send only the reduced set into Clay as context. Pulling everything raw into the LLM will always be costly