When something happens in the real world, your workflow fires — at the bar you set.
Evinor is a structured-events engine for the real world. It turns noisy news, filings, and press into typed, deduplicated events — each carrying a confidence score, a corroboration count, and typed sources. Subscribe to what matters, decide what's strong enough to act on, and get one event pushed to your webhook or Slack the moment it clears your bar. No scraping, no polling, no code.
One clean event — on your webhook, or in Slack.
Not an article to parse. Automations get a typed JSON record they can branch on; people get a readable Slack card. Same event, same instant.
Built for the next node in your flow.
Filter on type, gate on confidence and sources, map entities[].roleinto your CRM. It's plain, predictable JSON your automation can gate on — confidence, source count, source type — not scraped text it has to interpret.
// one event · one delivery · signed { "type": "Acquisition", "occurred_at": "2026-06-28T14:20Z", "entities": [ { "role": "acquirer", "name": "Databricks" }, { "role": "target", "name": "Tabular" } ], "deal_value_usd": 1000000000, "confidence": 0.97, "sources": 12, "watchlists": ["Portfolio"] }
"What if the event is wrong?" Inspect it before you act.
Every event arrives with its evidence attached — confidence score, corroboration count, signed payload. Your workflow can inspect the claim before it acts, and you can audit exactly what it acted on afterward.
// evidence attached · signed at delivery { "type": "Acquisition", "event_id": "evt_8f3a1c", "confidence": 0.94, "corroboration": 12, "sources": [ { "type": "official_filing", "n": 1 }, { "type": "newswire", "n": 4 }, { "type": "press", "n": 7 } ], "signature": "ed25519:9f2a…c71" }
Your workflow fires once, not forty times.
Forty outlets cover the same acquisition. Without dedup, that's forty webhook calls, forty Slack pings, forty CRM writes. Evinor collapses them into one canonical event — so your automation runs exactly once.
Accuracy isn't a promise. It's a dial you hold.
Every event is graded — confidence, corroboration count, typed sources. Rules gate on source strength, not just event type. You decide what's strong enough to act on, per rule, per workflow.
Rumor to official record — one evolving object.
A story doesn't arrive fully formed. Evinor delivers it as a single event that gains strength over time — same event_id throughout. Subscribe at whatever strength your workflow deserves.
rumor → corroboration → official record · no incumbent delivers this as one subscribable object
Same events, different bars.
Three ways teams wire Evinor events into what they already run — ordered by cost of error. Each fires at a different bar: the cheaper the mistake, the looser the gate.
Works in your stack today.
If your tool speaks webhooks, it already speaks Evinor. Point a signed webhook at n8n, Zapier, or Make and start building right now.
The deeper story, one layer down.
If you're building agent loops, the primitives matter. Evinor is a push-based event source with canonical resolution and a stable, graded schema — designed to be sensory input an agent can reason over, with its own confidence attached.
Push, not poll
Agents subscribe once and receive events the instant they exist — no polling loops burning tokens and latency to discover nothing changed.
Canonical entity resolution
"DBX", "the SF lakehouse firm", and "Databricks, Inc." resolve to one stable entity your agent can key on across every event.
A graded schema you can gate on
Discriminated type, versioned fields, plus confidence and source counts on every event — input your agent can gate on: confidence, source count, source type, instead of re-parsing prose.
Under the hood this is content-addressed messaging — but you don't have to think about it. Today Evinor covers business incidents — acquisitions, funding, insolvency, layoffs, recalls — and the engine generalizes to new event types and entities over time.
See what a rule would have caught — including what you found out about late.
Define a rule and run it against the last 90 days of real history before you turn it on. Every match it would have fired — and every one you heard about days after it mattered.
This is real — try it liveWire the real world into your workflow.
Request access and we'll help you define your first rules and pipe clean events into the tools you already run.