RSS feeds, alongside your mail

inboxy can follow your RSS / Atom feeds and surface the articles through the same agent connection and API as your mail. One login, two streams — your @inboxy.net inbox and your reading list, side by side in Claude (or any MCP-aware client).

Adding a feed

Account → Feeds → paste the URL → Add. Optionally give it a label so you remember what it is. From an agent:

feed_add({ url: "https://aws.amazon.com/blogs/machine-learning/feed/" })

Don’t worry about exact form — https://Foo.com/feed/ and https://foo.com/feed/?utm_source=x are recognised as the same feed, so you won’t end up subscribed twice.

How often inboxy checks

Every feed is polled once an hour, on inboxy’s behalf for everyone subscribed — so the publisher sees one polite request, not one per subscriber. If a feed starts erroring, we back off (up to once a day) until it recovers; after a long streak of failures it’s marked error in your list.

Can’t wait an hour? Hit Refresh on the feed. That’s rate-limited so nobody can hammer a publisher, but a successful refresh benefits every subscriber.

Reading articles

Today, through your agent or the REST API:

• From an agent: list_articles, get_article, summarize_article, summarize_feed.
• From a script: /v1/articles and /v1/feeds/:id/articles on api.inboxy.net.

An in-app reader is on the wish-list — drop us a line if you’d use one.

Safer link clicks. Every link in an article body is routed through inboxy’s redirector (l.inboxy.net) — the same one that strips trackers from forwarded mail. The publisher sees a request from inboxy, not from you, and if a link turns nasty we have one place to disarm it.

Summaries stay private. When you ask your agent to summarise a feed or article, the LLM running in your client does the work — inboxy never sees or stores the summary.

Privacy

Your subscriptions and read state are yours alone. inboxy will never tell one user that another follows the same feed — it’s not exposed in any tool, API response, or UI.

If you’re the last subscriber to a feed and you remove it, the shared cache is cleaned up after a 7-day grace period.

Limits

• Early Adopter plan: 10 feeds.
• Paid plans (coming soon): 50 feeds.

You can’t set a custom polling cadence per feed — every subscriber shares the same hourly check, and the adaptive back-off keeps things sensible.

Not yet supported

• Articles as inbox messages — ask if you want it; it’s a small change but we’d like to hear a real use case first.
• OPML import / export.
• WebSub push (PubSubHubbub) — everything is polled today.
• A built-in web reader. Use your agent or the API for now.