AutoXiv is open infrastructure: anyone can submit. Labs and companies use it to make their releases verifiable — a pinned repository, an independent sandboxed run, and a claim ledger that readers and agents can inspect instead of taking the PDF on faith.
Register the repository at the exact commit your paper describes. A sandboxed agent installs and runs it, typed claims are extracted and blind-confirmed, and the scorecard is published next to the paper — clamped to what the sandbox actually observed.
See the reproduction agentRun the same linter our agent uses on a draft and its repository: are the claims numeric, is there a run command, do the numbers in the results file match the paper, is the environment pinned. A 0–100 score with concrete fixes, before anything is public.
Check a draftOne MCP endpoint serves papers, overviews, reproduction records, and claims to Claude, Cursor, ChatGPT, or your own agents. Read-only works without signup; a key unlocks writes, pipelines, and a usage dashboard.
Read the MCP docsPrivate workspaces with roles and invitations, lab-scoped agents, and chat grounded in your own papers and reproduction history. Public directory pages for labs that want to be found.
Open workspacesAlready on arXiv? Import your lab’s papers in bulk — search by author, tick what is yours, register in one go.
Submitting, hosting, overviews, and reading are free and will stay free. Verified reproduction of your releases, larger agent quotas, and private workspaces for teams are arranged with each lab individually — tell us what you publish and we will tell you what we can verify.
Talk to usReproduction is automated for computational work with a runnable repository. We cannot verify wet-lab results, and a run that executes and matches the reported numbers is evidence — not proof — that a result is correct. Every verdict shows what was observed and what was not.