Summary
Put something in, find it again by meaning.
This is the smallest useful shape for a vector store: write and hybrid search, nothing around them. Where it points matters more than what it does — the default address is an instance the project hosts rather than a database of yours, so the first decision is whether your objects should be going there at all.
What it is
A thin client for a Weaviate vector database, installed from npm and started by your MCP client. Two tools: one puts an object in, the other searches.
What you get
- Objects inserted into the vector store one at a time
- Hybrid search over the store — keyword matching and vector similarity in the same query
Requirements
An API key for the Weaviate instance it talks to. The address it uses defaults to a hosted endpoint the project runs, and there is a setting that redirects it at a deployment of your own.