Turn your paper into a claimset

You have written a paper, perhaps already published. Your AI agent can turn it into a claimset: the same research as separate claims, each with its evidence, conditions and limits, which you check and publish here. It takes your agent the work and you the checking. Only the paper's authors do this.

Why bother

• AI assistants read it right. People increasingly meet research through an assistant. Given a PDF, it summarises what it guesses matters; given a claimset, it finds each of your findings with the conditions it holds under and what it rests on, and cites it exactly (entrylib:2026.000001@v1#C3).
• Each main claim has its own page and score. Every principal claim gets a page and a Support score showing how fully the paper's record backs it, and what would back it more (how).
• It is found where research is now searched. Claimsets are listed for search engines and answer engines, and any assistant can search and read them through EntryLib's MCP server.
• It stays yours and stays current. You are its responsible researcher. A correction is a new version, every version is kept, and the claimset links to the published paper. Publishing is free, with no account beyond your ORCID iD.

How

1. Give your agent the paper (the published or accepted version, with its supplementary material) and the message below.
2. Answer its questions: that you are an author, which author answers for the claimset, its licence, which findings matter most.
3. Check what it gives you: every claim, with where in the paper it comes from. This is the part only you can do.
4. Approve the short plain-language summary it drafts; it is shown first on the paper's page.
5. Let it submit, then confirm by signing in with ORCID within 48 hours. See Before you submit.

The message for your agent

I am an author of the attached paper and want to publish it on EntryLib (https://entrylib.org) as a claimset.

Follow the claimset-from-paper skill. If you can run commands: pip install claimset, then claimset guide claimset-from-paper, and the skills and specification it names. If you cannot: read https://github.com/entrylib/claimset/blob/main/skills/claimset-from-paper/SKILL.md and what it names.

- Ask me its questions before you build anything.
- Use only what the paper says: add nothing, drop no condition, and keep results where the method did worse.
- Draft a plain-language summary of at most 150 words for me to approve.
- Before submitting, give me every claim with where it comes from in the paper, so I can check it.
- Validate the file, and submit only when I say so.

Agents that can run commands (Claude Code, Codex, Cursor and others) do all of it. A chat assistant without them can still write the claimset; connect EntryLib's MCP server so it can check and submit it, or upload the file it gives you at Submit, which checks it first.

Not an author? A paper by others is not submitted here.