search_reference
Search academic literature databases (DBLP, Semantic Scholar, arXiv, OpenAlex) to find research papers and return properly formatted BibTeX citations.
LABSCO SUMMARY
Four sources in one query, and no key to get started.
Hitting all four indexes in parallel is what makes this worth a tool call instead of a browser tab: preprints on arXiv, conference records in DBLP, and the wider coverage of Semantic Scholar and OpenAlex resolve together. No account is needed, and the only key involved is optional and buys you throughput, not access.
What it is
A reference-search server that queries arXiv, DBLP, Semantic Scholar and OpenAlex concurrently and generates BibTeX from what it finds. You run it locally from a clone of the repository, launched with uv.
What you get
- One search across arXiv, DBLP, Semantic Scholar and OpenAlex, run in parallel rather than one after another
- BibTeX generation for the matches, ready to drop into a bibliography
- A local stdio server that uv starts from your clone
- An optional SEMANTIC_SCHOLAR_API_KEY that raises your Semantic Scholar rate limit
Requirements
Python 3.11 or later, uv, and a clone of the repository; the server is launched with uv run --directory <YOUR_PATH_TO_BIBTEX_MCP> python -m reference_mcp --stdio. A Semantic Scholar API key in SEMANTIC_SCHOLAR_API_KEY is optional and raises your rate limit.
Setup effort
Build from source — git clone https://github.com/xeophon/bibtex-mcp.git
USE CASES
01
Citing a paper you only half-remember
Search by title fragment or author and get a BibTeX entry back rather than hunting for the right DOI.
02
Filling a bibliography in one pass
Work through the reference list in conversation, collecting entries as you go.
03
Checking whether a preprint has been published
One query surfaces both the arXiv entry and any DBLP conference record.
04
Keeping searches off a third-party host
Run the server locally over stdio, so searches go from your machine straight to the indexes.
4 tools
Listed in project code.
search
Search academic literature and return lightweight results for browsing.
fetch_multiple
Fetch full documents for previously searched references.
fetch
Fetch a single document by ID.
BEFORE YOU INSTALL
What to know first
Tools you get
4 tools registered in the project's code. The README names no tools. These 4 are the functions reference_mcp/server.py registers as tools; fetch is a single-ID version of fetch_multiple kept for ChatGPT. Nobody has connected to it, so 4 is the source's count, not a measurement.
Requires
- uv on PATH
- Python 3.11+
Where it runs
The documented client runs the server on your machine.
INSTALLATION
Connect your client
You clone this one yourself.
Server setup
TERMINAL
git clone https://github.com/xeophon/bibtex-mcp.gitIn the configuration under MCP Clients, <YOUR_PATH_TO_BIBTEX_MCP> is the absolute path of the cloned bibtex-mcp folder. SEMANTIC_SCHOLAR_API_KEY is optional; leave it out if you have no key.
MCP Clients
Any MCP client — uv
ANY MCP CLIENT — UV · JSON
{
"mcpServers": {
"bibtex": {
"command": "uv",
"args": [
"run",
"--directory",
"<YOUR_PATH_TO_BIBTEX_MCP>",
"python",
"-m",
"reference_mcp",
"--stdio"
],
"env": {
"SEMANTIC_SCHOLAR_API_KEY": "<YOUR_SEMANTIC_SCHOLAR_API_KEY>"
}
}
}
}
