Getting started
Introduction
What EdgeOrigin is, who it's for, and how the real-time data is delivered.
Overview
EdgeOrigin is the real-time data layer for markets and machines. Our own global edge network listens to real-world events as they happen; our machine learning models evaluate each one for urgency, authenticity, and data quality and turn it into structured signal; and we deliver it over one API and streaming interface.
Every record is timestamped at the edge and carries its provenance, so you can act on it and trace it back to the observation behind it.
Who it's for
Two audiences - markets and machines. Hedge funds and quant desks consume real-time, ground-truth signal to trade on; AI and ML teams consume large, fresh, high-signal datasets to train and ground models.
Tip · One dataset, two ways to consume it
How it works
Listen, evaluate, deliver. Our edge nodes hear events in real time, machine learning deduplicates the stream and evaluates urgency, authenticity, and data quality, and you subscribe to the output. Authenticate with an API key, then pull records or open a streaming connection:
GET https://api.edgeorigin.com/v1/records?stream=<stream>&since=<ts>
Authorization: Bearer <API_KEY>What a record contains
A record is not only a payload. Each one carries the evidence for its own timing, origin, and trustworthiness, because those are the three things you will be asked about after you act on it:
Timing. observed_at is the instant the edge node heard the event, to the millisecond, against a network-wide reference clock. Because every node shares that clock, records from different regions are directly comparable and ordering holds globally - which is what makes a backtest on our history honest.
Origin. heard_by names the node that received the event and served_by names the point of presence that delivered it to you. Together they let you recompute end-to-end latency per record instead of trusting an aggregate figure.
Trust. urgency, authenticity, and data_quality are the values our models assigned in flight, exposed as numbers so you set your own thresholds. provenance carries signed lineage - source class, corroborating observations, and the evaluation trail - so a record cannot be altered between our network and your systems.
The architecture behind those guarantees is described on the edge network page.
Choosing streaming or bulk
Both read the same records through the same authentication, so this is a question about your workload rather than about the data. Open a streaming connection when the value of a record decays in seconds and you want it pushed the moment it is observed - a live desk, an alerting path, a model that has to stay current. Use the REST endpoint when you want a defined range of history, a backfill after downtime, or a reproducible dataset for a training run.
Most teams use both, and that is the intended shape: research reads history, production reads the stream, and because they are the same records with the same evaluation values attached, what you validated is what you deploy. Prefer a long-lived stream to frequent polling - polling adds an interval of delay you cannot recover and a handshake you pay for repeatedly.
Coverage and freshness
Coverage is organized by subject area rather than by source, because the sources behind an area change and your integration should not. EdgeOrigin covers the AI era and the physical systems underneath it: AI and compute, data centers, chips and fabs, nuclear and energy, critical minerals, space and satellites, networks, industry and robotics, markets, filings and disclosure, and conflict and hazards.
Freshness is a property we publish rather than imply. Because every record carries the timestamp of the node that heard it and the node that served it, you can measure how current any subscription is at any moment, and a region degrading is visible in your own data rather than something you discover from a status page.
Transport security
Both the streaming and REST endpoints are served over TLS 1.3 with hybrid post-quantum key exchange, so a session recorded in transit today cannot be decrypted later. Details and the standards involved are on the post-quantum security page.
Next steps
Continue to the Quickstart to pull your first records and subscribe to the stream in minutes.
Was this page helpful?