Last updated: 1 November 2026
Paper Q&A is a Chrome extension that lets you ask questions about a PDF and highlights the answers. This policy explains what data it handles. The short version: the extension has no server of its own and collects no analytics. It sends the paper's text and your questions only to Anthropic's API (or, if you choose, to your own Claude Code installation), and only when you press Find answers (or, in the Implement tab, Write implementation prompt). When you point at or click a citation, or click the paper info button, it looks that one paper up in public scholarly databases (Semantic Scholar, Crossref, OpenAlex and arXiv).
All of the following stays in your browser's extension storage (chrome.storage.local) on your device:
- Your Anthropic API key, which you enter on the options page.
- Your settings: whether to use the Claude API or Claude Code and the folder for Claude Code projects, chosen model, effort level, what to do when a PDF opens, whether citation previews are on, and your Semantic Scholar API key if you added one.
- A cache of your questions and Claude's answers for each PDF, keyed by a SHA-256 hash of the file. This lets a paper's highlights come back when you reopen it.
- A cache of implementation prompts from the Implement tab for each PDF, keyed the same way: the systems Claude found in the paper, the prompt it wrote, and the notes you typed for it.
You can clear the cache on the options page. Removing the extension deletes all of it.
When you press Find answers (or Re-ask all), the extension sends this data to Anthropic's API at https://api.anthropic.com:
- the text extracted from the PDF currently open in the viewer,
- the questions you typed,
- your API key, which authenticates the request.
When you press Write implementation prompt in the Implement tab, the extension sends the same paper text and API key, plus the notes you typed for the prompt and the name and summary of the systems Claude found in the paper.
If you tick Search the web for a question, Claude may also search the web for it. In the Implement tab, Search the web also lets Claude fetch web pages, such as the paper's official code repository, by their address. The searches and fetches are run by Anthropic as part of the same API request, so the extension itself still contacts only api.anthropic.com. The search queries and page addresses Claude chooses, which can draw on your question and the paper, are handled under Anthropic's terms. Links to the web pages Claude cites are shown in the sidebar; the extension does not open them unless you click one.
If you choose the Claude Code backend on the options page, none of the above goes to api.anthropic.com from the extension. Instead the extension passes the same data (without an API key) to a helper program you installed on your computer, which runs your own Claude Code installation with it. Claude Code sends it to Anthropic under your Claude Code login. When the Implement tab starts Claude Code on a prompt, the helper also writes the prompt and a copy of the PDF to a new folder on your computer, in the folder you chose on the options page. The helper runs only on your computer and sends nothing anywhere else.
Anthropic processes this data under its own terms and privacy policy, which apply to your API account: https://www.anthropic.com/legal/privacy. Paper Q&A is an independent project and is not affiliated with or endorsed by Anthropic.
When you point at or click a citation in the paper, or click the paper info button in the viewer's toolbar, the extension looks up that paper's details. The requests go to these public services:
- the Semantic Scholar API (
https://api.semanticscholar.org, run by the Allen Institute for AI): the paper's identifier or title, plus your Semantic Scholar API key if you added one; - the Crossref API (
https://api.crossref.org): the text of the clicked bibliography entry (authors, title, venue, year), or the paper's title; - the arXiv API (
https://export.arxiv.org): the paper's title; - the OpenAlex API (
https://api.openalex.org): the paper's DOI.
The identifier is a DOI, or an arXiv or ACL Anthology ID. The title comes from the bibliography entry, or from the open PDF's metadata or first page. No other text from the PDF is sent, and nothing is sent unless you point at a citation, click one, or click the info button. These services' own terms apply: Semantic Scholar, Crossref, arXiv and OpenAlex. The papers found are cached on your device for a week; Clear cache on the options page removes them. You can turn citation previews off on the options page. Links in the card (Google Scholar, arXiv, DOI and so on) are opened only if you click them.
The extension also downloads the PDF you choose to open, from the web address it was opened from. This is the same request your browser would make to display it.
Nothing else is sent anywhere. In particular, the extension does not send your browsing history, the addresses of pages you visit, or any data about you to the developer or to anyone else.
Chrome asks you to allow Paper Q&A to "read and change data on all websites". The extension uses this permission only to:
- notice when a tab is showing a PDF, by its address or its
application/pdfcontent type, so it can offer to open it; - show the small "Open in Paper Q&A" banner on that PDF tab;
- download the PDF into its viewer, including from sites that need your login cookies and from local files if you allow file access;
- if you choose to always open PDFs in Paper Q&A, send a PDF tab to its viewer as soon as the PDF starts loading.
It does not read or change the content of any other page.
The developer receives none of your data, so there is nothing to sell or share. Data is not used for advertising, credit decisions, or any purpose other than answering your questions and showing the papers you click.
Paper Q&A is not directed at children under 13.
Changes to this policy will be published in this file in the project's repository, https://github.com/mohph197/pdf-qna, with a new date at the top.
Open an issue at https://github.com/mohph197/pdf-qna/issues.