Claude research mode on Zotero library + 10GB storage free w/ Cloudflare
Hey folks! Just popping in the share a (pre) weekend open-source project…
I’ve been working with larger libraries (10k+ records) recently and have ended up rolling my own Zotero sync + full text search solution backed by Cloudflare. It gets you 10GB of cloud storage for free (under $20/mo for 1TB of PDFs) & exposes a remote MCP so Claude can actually use your library as a source in research mode.
All ive done is host a thin WebDAV <> S3 server using my homelab and tailscale. Then, on the same machine, I have a worker that ingests new PDFs (md conversion -> chunks -> embedding) and stores the vectors back in the same bucket with LanceDB.
Put a remote MCP server on there too which then you can use to connect the whole thing to Claude. Since, it’s a remote MCP server, the research mode can actually use it as a source. See attached picture.
Repo link: https://github.com/ElliotRoe/lit-bucket