Fresh context for AI agents
Agents reason over what you feed them, and most pipelines feed them a stale nightly crawl. Hypeline turns the sources you care about into a live, deduplicated stream of new content, so your models act on what happened minutes ago instead of last week.
# one SSE call = a live tool for your agent
curl -N "https://api.hypeline.io/v1/stream?q=competitor+OR+launch" \
-H "Authorization: Bearer sk_live_..."
# each event: new content, extracted, deduplicated
{ "title": "Rival launches EU tier",
"url": "https://rival.io/blog/eu",
"emitted_at": "2026-07-13T08:12:41Z" }
# -> embed it, index it, or hand it to the agentFrom sources to agent context
Point it at your sources
Feeds, plain web pages, push sources: any public URL your agent should know about, in any language or region. Hypeline auto-detects the format.
Filter to what matters
Keyword and Boolean matching runs server-side before delivery, so your agent receives the slice it cares about instead of a raw firehose of everything.
Feed your agent live events
One SSE call opens the stream. Each event is new content, already extracted and deduplicated, with a source URL and timestamp your agent can cite.
Why agents deserve better than a nightly crawl
| Scheduled crawl | Hypeline stream | |
|---|---|---|
| Freshness | Hours to days old by the time it lands | Push sources in seconds, feeds and pages in minutes |
| Duplicates | The same page over and over, every crawl | One event per genuinely new piece of content |
| Noise | Boilerplate, ads, and timestamps in your context window | Extracted main text only |
| Cost shape | Re-embed the whole corpus per crawl | Embed only what actually changed |
| Provenance | A snapshot with no clear origin time | Source URL, emitted_at, and an Ed25519 signature per event |
| Recovery | Missed a run? That window is gone | Cursor catch-up replays exactly what you missed |
Frequently asked questions
How do I connect an agent or RAG pipeline to Hypeline?
Over plain HTTPS. Open the SSE stream with one request and a bearer key, and treat each event as a document to ingest: extract, embed, and index it, or hand it straight to the agent as context. If you prefer push, HMAC-signed webhooks deliver each event to your endpoint.
Why not just crawl the sites myself on a schedule?
A scheduled crawl re-fetches everything, re-embeds mostly unchanged content, and is stale between runs. Hypeline watches the sources continuously, extracts the main text, and emits only genuinely new content, so your pipeline processes changes instead of snapshots.
Can my agent trust where an event came from?
Each event carries its source URL and an emitted_at timestamp, and every event is signed at the source with Ed25519. Your pipeline can verify the signature with a public key before acting on the content.
What happens if my consumer goes down?
Nothing is lost. Events are durably committed before delivery, and the stream is resumable: reconnect with your last cursor and you receive exactly what you missed, in order.
How fresh is the content, honestly?
Latency depends on the source tier. Push sources arrive in seconds, feeds in minutes, and watched web pages in minutes to tens of minutes. We publish it per tier rather than promising one universal number.
Be first to make the static web live
Add your email and we'll tell you the moment it goes live.
No spam. One email when it goes live.