Magpie is an unofficial index of Claude artifacts that people published to the open web. This page says what it holds, where that came from, and how to get it out.
For each artifact the index stores its public URL and ID, the title and description as they appeared on the page that linked to it, a category and tags derived from that text by keyword rules, the address of the page the link was found on, and the date it was recorded.
None of this is written by us. It is copied from material the artifact's author chose to publish, and from public pages that linked to it. If a title or description contains personal information, that is because it was in the published artifact.
A crawl runs once a day through Firecrawl over public search results. It reads pages that mention a claude.ai artifact link and extracts the address from the text. It does not sign in to Claude, does not read private artifacts, and does not copy the contents of an artifact.
To run a searchable catalogue of material that is already public. If you are in a place with data protection law that asks for a lawful basis, ours is legitimate interest in indexing published material, balanced against your right to have a record removed on request, which the next section covers.
Use the removal form. You sign in, paste the artifact link, and the record is deleted from the index. The artifact ID goes on a blocklist so the daily crawl will not add it back.
Removing a record here does not unpublish anything on claude.ai, and it does not reach Google, the Internet Archive, or anyone who copied the link. Unpublish at the source first, then remove the record here.
Sign-in uses Google or GitHub through Supabase. We receive your user ID and email address from the provider, and nothing else: not your contacts, not your repositories, not your password.
An account is needed for two things, and only these two:
Putting your name to a removal request, so there is a record of who asked. And holding a Pro workspace, where we store the artifact IDs you saved, the searches you saved along with when you last opened them, and, if you bought Pro, a row recording the Paddle transaction ID, the price ID, the amount and the currency.
Reading the index needs no account and never will. We do not send marketing email, and there is no mailing list to be on.
Paddle.com Market Ltd is the merchant of record. Card details go to Paddle and never reach this site or its database. Paddle collects what it needs to take a payment and charge the right tax, including your billing country, under its own privacy policy, and it is the data controller for that.
What comes back to us is the transaction ID and what was bought, which is what the entitlement is based on and what makes a refund traceable.
Email support@weekndlabs.com from the address you signed in with. The account, its collections, its watchlists and any purchase record are deleted. It is done by hand, so allow a few days, and you get a reply confirming it.
Deleting your account does not remove artifact records from the index, because those are not yours and were never tied to your account. Use the removal form for that, before you delete the account you would need to sign in with.
Vercel Web Analytics counts page views. It sets no cookies, does not follow you to other sites, and identifies a visit through a hash that is discarded and regenerated daily, so there is no profile to build. We see counts per page, not people.
We also count what happens on the site in our own database: a category page opened, a locked card clicked, a sign-in started, a checkout opened. Each row holds the name of the thing that happened, which tier you were in, the page path without its query string, and a random value kept for that browser tab only. No IP address, no account id, and never what you searched for.
That random value is held in sessionStorage rather than a cookie, so it disappears when you close the tab and cannot follow you to another visit. It exists to tell whether one person clicked three things or three people clicked one thing each, which is the whole of what these counts are for.
There is no advertising and no other third-party tracking. Browser storage is your light or dark theme choice, that per-tab value, and a Supabase session token if you sign in.
Records are held in a Postgres database hosted by Supabase. The site runs on Vercel. Firecrawl performs the daily crawl, and Vercel Web Analytics counts page views. Payments go through Paddle. All of them may process data outside your country.
Magpie is run by WeekndLabs. For anything this page does not answer, to ask what we hold on you, or to object to our holding it, email support@weekndlabs.com.
To take a single artifact record out of the index, the removal form is faster than an email and needs no reply from us.
This is a plain description of how the site works, not legal advice.