Privacy policy

What we keep

Your Google account data: Google identifier, email, name and profile picture. New accounts only use Google sign-in. If you created an account with an email and password in the previous version, we keep a hash of that password (never the password itself) until you delete the account. If you can no longer sign in because your email isn't a Google one, write to hello@matcher.news from that email and we'll delete it.

Data from the previous version of the service (the news reader): the articles you saved or marked as read, the username of your public profile, the Nostr authors you followed or muted, the newsletter subscription, feedback messages and anonymous visit statistics remain in the database until a scheduled cleanup. We no longer collect new data of this kind. If you delete your account, everything associated with it is deleted; the anonymous visit statistics aren't linked to any account.

Your feeds: the URL of each feed, its title, the site's URL, the date you added it, the folder you organized it in and the last error when reading it. Folders from the previous version are kept as your folders.

Your connected agents: the name each agent declares, its return addresses and the date it was last used. Access tokens are only stored as a hash.

MCP server usage: for each call we keep which tool was used, which agent made it (the OAuth client identifier), whether it worked, the error code, the latency, the number of results and the date. We don't keep the arguments or the content of the responses.

For “what's new” to work, we temporarily keep which articles we already delivered to each agent (14 days) and an index with the title, link and date of those articles (7 days).

Sessions and security: session cookie, IP address and browser of each session, and security logs to prevent abuse.

Who processes the content

Matcher News reads the public feeds you add and delivers the headlines to your agent and, when the feed publishes it, the full text.

We keep a 30-day archive of what the feeds our users follow publish: title, excerpt, link, date and, when the feed itself publishes the full text, that text (up to 20,000 characters per article). It is shared per feed, not per user, and lets your agent search and read articles that have already left the feed. We only keep what the feed publishes, never text extracted from pages. When a feed's address looks private (credentials, a token in the address or a private-feed service we know, such as a paid newsletter's private feed) we keep the title and excerpt, not the full text; this detection is automatic and may not recognize every case: if your private feed isn't recognized, write to hello@matcher.news. What is kept from a feed is visible to the agents of everyone who uses that same address, including a private feed's. A publisher can ask us not to keep its content by writing to hello@matcher.news: we stop keeping it and delete what was kept within 24 hours.

When your agent asks for an article whose feed only has an excerpt, Matcher News may download that article's public page (the link the feed published, never an address sent by the agent), extract its text and hand it only to your agent in that response. We keep neither the page nor the extracted text. Our reader, MatcherNews-Reader, does this; it honors robots.txt and each site's noai/noindex directives, and publishers can opt out as explained at /bot. Your agent can skip it by asking for the article with fetch_page: false.

The content is processed by your agent (for example Claude, ChatGPT or Cursor) under that provider's privacy policy. We don't control what the agent does with it.

This site doesn't show news: it only lets you manage your feeds and connect your agent.

Cookies and measurement

We use a technical session cookie and, only if you pick a language with the language button, a cookie that remembers it (otherwise we use the language set in your browser). When you sign in with Google we also use two technical cookies that last 10 minutes, to validate the sign-in and remember which page to return to.

On the public pages of this site we measure visits (page views only) with Google Analytics, which uses its own cookies, and with Vercel Analytics. They aren't used in My feeds, the Catalog, the OAuth consent or Connections.

Providers

We host the service on Vercel, the database on Neon (Postgres) and the cache on Upstash (Redis). Google handles sign-in. We don't sell your data or use it for advertising.

Deleting your account

You can delete your account at any time from your account menu. Deletion is final and immediate: it deletes your account, your feeds, your sessions and your agents' tokens.

The temporary “what's new” state is deleted at that moment; if the cache doesn't respond, it's retried within minutes and, in the worst case, it disappears when it expires (7 to 14 days).

The 30-day archive of what feeds publish is shared by everyone who follows each feed. If you were the only one following a feed (for example, your newsletter's private feed), what was kept from it is deleted by the automatic daily cleanup, normally within 24 hours of deleting your account or removing the feed (if that cleanup fails one day, on the next one).

MCP usage records are unlinked from your account: your user identifier is deleted and they only keep the pseudonymous identifier of the agent that made each call.

You can disconnect an agent without deleting your account in Connections; it loses access in under a minute.

Contact

If you have questions about privacy, write to us at hello@matcher.news.