Web search returns abstracts, paywalls, and citations that cannot be checked. Redpine returns the relevant chunk of the document.
Redpine licenses full-text research directly from publishers and research institutions, and serves it to ChatGPT/Codex over MCP alongside a large open-access corpus. One query searches both.
Provenance by default. Every result carries its publisher, title, publication date, and a resolvable identifier such as a DOI. Licensed content and open-access content are labeled separately, so the two are never confused inside an answer.
You see the price first. Any search can be previewed for free. ChatGPT/Codex sees the title, source, and a snippet of every result, plus the exact cost to unlock them, then asks before spending anything. Unlock the three results that matter instead of the 30 that came back. The same results re-fetch free for seven days. Discovery, schema inspection, previews, and balance checks are always free.
Built for questions where being wrong is expensive. Clinical decisions, systematic review, research engineering, and any analysis that has to survive a source check.
An account can also hold data integrations beyond literature. ChatGPT/Codex reads what that account is entitled to at runtime rather than assuming a fixed tool list.
Compensation flows back to the rights holders. New accounts get free queries. Credits are bought on the Redpine dashboard, and the connector never handles payment details: it shows the price and asks before anything is charged.
Requires a Redpine account. Sign-in is OAuth on first use.