We had 400 feeds, 40 newsletters and an intern no time to read any of them. So we built something that reads them for us.
Find every feed. Scrape the sites that have none. Deliver only the stories that matter: scored, summarized and ready to publish.
Runs on Cloudflare's edge · robots.txt respected · excerpts and links, never full articles
Live · The pipeline in motion
A software provider used by multi-location dental groups disclosed unauthorized access to a support environment. The company says patient-record systems were not reached; affected clinics are being notified.
Why it mattersClinics that grant the vendor remote support access should rotate credentials and review access logs from the disclosure window.
Models score and summarize. Editors approve what ships.
Illustrative example; the vendor and story are fictional.
The three stages
01 / Feed finder
Every day, several AI models search the web for sources on your beat. EOXMIND visits each site and maps every machine-readable way in: RSS and Atom, JSON Feed, sitemaps, public APIs. Then independent judges score each one against your brief.
02 / Scraping
Listing pages and watched pages are fetched cheapest-first: a conditional request that costs nothing when nothing changed, plain HTML next, a real browser only when a page needs JavaScript. When a site redesigns, EOXMIND proposes new selectors for one-click approval.
03 / Delivery
Free checks run first: your keywords, duplicate and near-duplicate detection. A cheap model scores what's left against your editorial brief; only the best gets a full summary. Approved stories go out as signed webhooks, a JSON Feed and a daily edition.
An illustrative day for one publication. Your ratios depend on your sources and brief.
Other tools hand you a firehose, or let an AI publish on your behalf. EOXMIND does neither. Models sort, score and summarize; a person approves what your readers see, and every approval or rejection teaches the triage what you care about.
One pipeline. Two front doors.
// developers
// api/eoxmind.ts: every signal arrives signed
export async function POST(req: Request) {
const body = await req.text();
const [t, v1] = parse(req.headers.get('x-eoxmind-signature'));
if (now() - t > 300) return reject();
if (!equal(v1, hmac(secret, `${t}.${body}`))) return reject();
await publishDraft(JSON.parse(body));
}
// or pull it
GET /v1/ddsbreach/feed.json // JSON Feed 1.1Signed webhooks, a pull API and JSON Feed. No SDK required.
// editors
Clinics offering moderate sedation must renew permits under the new schedule by next spring.
A review queue, not a firehose. No code required.
Pipeline receipts
These are design limits of the pipeline, not benchmarks. Real savings depend on how often your sources change.
Compared to the usual way
| Capability | EOXMIND | Manual monitoring | Keyword alerts |
|---|---|---|---|
| Finds feeds, sitemaps and APIs for you | Yes | No | No |
| Watches sites with no feed | Yes | By hand | Partial |
| Polite crawling (robots.txt, pacing) | Yes | n/a | Varies |
| Merges duplicates across sources | Yes | By hand | No |
| Scores relevance against your brief | Yes | You do | Keywords only |
| Delivers to your site | Signed webhook + feed | Copy and paste | |
| One scraper, many publications | Yes | No | No |
A comparison with common workflows, not with any specific product. Spot something unfair? Tell us.
Point EOXMIND at your beat. It finds the feeds, watches the rest, and hands you only the signal.