AutoXiv hosts research papers with a pinned repository, a structured AI overview, and an agent-run reproduction record. Free to submit. Readable by people and by agents.
Already on arXiv? Register it by link →
Upload a PDF, or paste an arXiv link and AutoXiv imports the record and links the original PDF. Add the GitHub repository at a specific commit. No fees, no embargo, no queue.
A structured overview — problem, approach, results, limitations — is generated and the submission is screened before it goes live.
A sandboxed agent installs and runs the code, typed claims are extracted and blind-confirmed, and every verdict is clamped to what the server observed.
A question answered across every paper on the platform, grounded in their overviews and evidence, with citations that open the paper.
OpenBefore you submit: a linter that reads your PDF and repository and scores how reproducible the release is, with concrete fixes.
OpenClones a paper's repository into a sandbox, installs, runs, and compares the output with the paper's claims.
OpenPool stakes on papers that still need an independent reproduction. Claims link to verified runs.
OpenScreens a paper for tradable signals and drafts a strategy spec and backtest scaffold. Research tooling, not advice.
OpenBring a PDF and a repository. Your paper is live, summarized, and ready to be reproduced within minutes.
Lets Claude, Cursor, or any agent search, read, reproduce, and stake over MCP — no signup to read.