alphaXiv alternative (MCP)
Scholar Feed: an alphaXiv alternative built for keeping up, not just deep dives
Scholar Feed and alphaXiv are both MCP servers over arXiv, built for different jobs. alphaXiv is excellent for going deep on a specific paper — multi-turn agentic retrieval, PDF question-answering, and it can read the paper’s GitHub repository. Scholar Feed is built for keeping up with a field: a 600,000-paper CS/AI/ML corpus with an LLM novelty score on every paper, citation-graph lineage that reaches the current-month frontier, daily watches for new work, and an anonymous tier with no account required. Use alphaXiv to interrogate one paper and its code; use Scholar Feed to track a field and rank the firehose. Install with npx scholar-feed-mcp init.
Why people search "alphaXiv alternative for MCP"
alphaXiv is a well-funded, free research tool with a genuinely strong MCP server: it does agentic multi-turn retrieval over arXiv, answers questions about a paper’s PDF, and can even read the paper’s linked GitHub repository. It is one of the better tools out there for taking a single paper apart. Two things send people looking for an alternative:
- It is built around the deep dive, not the keep-up loop. If what you want is to track a lab, a technique, or a competitor over time and be told what is new, that is a different shape of tool.
- It asks you to sign in. alphaXiv’s MCP is gated behind an account, so anonymous, zero-setup access in a fresh Claude or Cursor session is not on the table.
Scholar Feed is the other shape: a curated CS/AI/ML corpus you point an assistant at, ranked for the firehose, with watches to keep you current. It is not better at deep single-paper reading — see below — it is built for a different job.
How Scholar Feed compares to alphaXiv
| Comparison axis | alphaXiv | Scholar Feed |
|---|---|---|
| Access shape | MCP server (account / sign-in required) | MCP server (anonymous tier, no account to start) |
| Coverage | ~2.5M arXiv (CS, math, physics, stats, quant-bio/fin) | 600,000+ CS/AI/ML, indexed daily from arXiv |
| Deep single-paper read | PDF Q&A + reads the paper’s GitHub repo | Extracts results/experiments from LaTeX source |
| Code exploration | Reads the linked GitHub repository | Flags code availability + repo link; does not read it |
| Per-paper signal | Paper content + Q&A | LLM summary + 0–1 novelty score |
| Keep-up | None (built for the deep dive) | Daily watches on a saved filter (lab, technique, author) |
| Recency | Agentic retrieval | Citation-graph cited_by reaches current-month work |
| Ranking | Relevance / agentic | Multi-signal (novelty, citation velocity, code, recency) |
Where alphaXiv wins (we are not going to pretend otherwise)
If your job is to take one paper apart, alphaXiv is excellent, and in our own hands-on testing it was the stronger tool at exactly that. Its PDF question-answering plus the fact that it can read the paper’s GitHub repository mean you can ask "how does this implementation actually differ from the equation in the paper" and get a grounded answer. Scholar Feed extracts a paper’s results and tables from LaTeX, but it does not read the code, and it does not do multi-turn interrogation of a single PDF. alphaXiv also covers more of arXiv — physics, math, statistics, quant-bio and finance — where Scholar Feed only knows CS/AI/ML.
What Scholar Feed is actually for
Scholar Feed is built for the part alphaXiv does not target: keeping up with a field and ranking the firehose. You ask your assistant "find recent high-novelty work on test-time compute scaling" and get CS/ML papers ranked by a novelty score and citation velocity, then "set a watch on new retrieval-augmented-generation papers above 0.5 novelty" and it surfaces matches daily. Under the hood it walks the citation graph, from a canonical paper out to the newer work that cites it, which is how it reaches 2026 results a model cannot recall and a keyword search buries. And every paper comes back with its real arXiv ID, title, and date attached, so citations are correct by construction.
The honest summary
These are not the same tool with a winner and a loser. alphaXiv is the better deep dive: one paper, its PDF, its code, across all of arXiv, if you are happy to sign in. Scholar Feed is the better keep-up: a curated CS/AI/ML corpus with novelty ranking, citation-graph recency, and daily watches, anonymously, with no account. Plenty of people will want both.
When NOT to use Scholar Feed
- You want to deeply interrogate one paper and its code. alphaXiv’s multi-turn PDF Q&A and GitHub-repo reading are purpose-built for that; Scholar Feed extracts full text but will not read the repo.
- You need physics, math, statistics, or quant-bio/finance papers. alphaXiv’s ~2.5M-paper arXiv coverage is broader; Scholar Feed is CS/AI/ML only.
- You specifically want agentic multi-turn retrieval over a single source. That is alphaXiv’s design center, not Scholar Feed’s.
Frequently asked questions
Is Scholar Feed a good alphaXiv alternative?
It depends on the job. alphaXiv is excellent for going deep on a single paper: it answers questions about the PDF and can read the paper’s GitHub repository, and it covers all of arXiv. Scholar Feed is built for a different job, keeping up with CS/AI/ML, with an LLM novelty score on every paper, citation-graph lineage that reaches recent work, and daily watches, all from an anonymous MCP server with no account. For deep single-paper reading, alphaXiv is the better tool; for tracking and ranking a CS/AI/ML field, Scholar Feed is.
Does Scholar Feed require an account like alphaXiv does?
No. alphaXiv’s MCP server requires you to sign in. Scholar Feed’s search and read tools work anonymously at 100 calls per day, no account needed. A free API key raises the limit to 1,000 calls per day and Pro raises it to 10,000. Install with npx scholar-feed-mcp init.
Can Scholar Feed read a paper’s code repository like alphaXiv?
No. Reading the paper’s linked GitHub repository is one of alphaXiv’s strengths. Scholar Feed flags whether a paper has code and links the repository, and it extracts the paper’s results and tables from LaTeX source, but it does not read the repository’s files. If interrogating the code matters, alphaXiv is the better choice.
Which one finds more recent papers?
Scholar Feed reaches recent CS/AI/ML work by walking the citation graph forward, from a canonical paper to the newer papers that cite it, which surfaces current-month results a language model cannot recall. alphaXiv uses agentic retrieval across a broader slice of arXiv. For the latest CS/AI/ML frontier specifically, Scholar Feed’s citation-graph approach is built for it.
Try it
npx scholar-feed-mcp initFree anonymous access is 100 calls/day (no account); a free key raises it to 1,000/day. Open source (MIT): scholar-feed-mcp on GitHub.
More setup options on the developers page.