How does Off-Nadir Delta tell me where to look and which sensor to use?

Start on the Watchfloor to see where activity is concentrated today — geolocated events (conflict, security, disasters, infrastructure), continuously refreshed, each with a severity score (ranked by imaging value on plans that include it). Then ask Delta Agent why a given event matters; it reasons over the live signals with real intelligence tradecraft and returns a sourced brief that recommends which satellite sensor and area to image — Sentinel-1 SAR for cloud, night, or vessel detection, Sentinel-2 optical for a visual look. Open the recommended scene on the Map to look closer. This "tip-and-cue" workflow takes you from the day's events to the right satellite image without switching tools.

Is there a free tier?

Yes. Create a free account to receive tokens and start exploring immediately — browse world events on the Watchfloor, ask Delta Agent, and view satellite imagery on the Map. No credit card required.

Do I need GIS experience?

No prior experience required. The interface is designed for beginners — navigate the map, search for satellite imagery, ask Delta Agent a question in plain language, and start exploring in minutes.

What satellite data is available?

Access Sentinel-1 C-Band SAR (works through clouds, day or night), NISAR L-Band SAR (NASA/ISRO — penetrates vegetation canopy; provisional calibration), Sentinel-2 optical (10 m resolution, 13 spectral bands), VIIRS Nighttime Lights (NASA Black Marble), and NASA FIRMS Active Fire hotspots (VIIRS 375 m, updated daily). No API key is needed for fire data.

Off-Nadir Delta

Event intelligence · AI-ready OSINT and satellite GEOINT

Off-Nadir Delta: From first signal to evidence.

Source-linked event intelligence for humans and AI.

Turn news reporting and open data into structured, geolocated events. Each one carries the reports behind it, a location only as precise as the reporting allows, and whether it could be seen from orbit. Follow developments and explore the satellite observations available—in the web app, over the REST API, or from your AI agent through MCP.

Seeing the event is easy. Knowing what you can prove—and what evidence is still missing—is the product.

Source-linked · Geolocated · Uncertainty-aware · AI-ready

Explore live intelligenceWatchfloor

Observe from spaceMap Track places & eventsWatchlist Read today's reportReports Explore the APIAPI & MCP

Livesignals ingesting

On every event

Not only that something happened: how well it is located, who reported it, and whether a satellite could see it.

Located, not just mentioned

Raw feeds place an event at country level, or at the city it was reported from. Delta resolves precise coordinates only as far as the reporting names the place, and drops a location that contradicts the reported country. Each event says how far it got: country, town, or the named facility. How events are located

Traceable, not asserted

Every event keeps its original sources and how many separate outlets carried it, and is updated as new reports arrive. The number of outlets shows how widely it is reported; it is not presented as confirmation.

Observable from orbit, or not

Each event says whether what was reported could be seen by a satellite, and why not when it could not. Some events come from a measurement — a heat source, ship counts, a seismic record — and say only what was recorded. Measured events

Trust & transparency

Built for trust. Designed for scrutiny.

Intelligence is only useful when you can examine where it came from, how it was processed, and what remains uncertain.

Traceable sources

Events link back to the reporting they came from. How sources are used, credited and attributed is written down. Sources and attribution

Transparent data practices

How source material is collected, how long it is kept, and what of it is shown are published as a policy you can read. Source & Content Policy

Explicit uncertainty

Where a location stops being precise, where reports conflict, and what a satellite could not show are recorded and stated, not smoothed over. Known limitations

It tells you when it doesn't know

Most of what happens cannot be settled from orbit — the place is too vague, or nothing physical changed. The product records which of those it is, per event, and says so.

Of the 2,353 events read in the last 72 hours, each one that cannot be looked at from orbit carries where it stopped — the location, or whether it would be visible from orbit — and why. These are the most common reasons, counted just now:

494What happened is visible only at about 0.3 m resolution (a single building), which needs the location pinned to a street block or facility; the reports give a coarser one. — stopped at visibility from orbit
271A report names a facility, but it could not be found on the map yet — it can be looked up. — stopped at location
218Nothing the reports describe would be visible from orbit. — stopped at visibility from orbit
214event_not_verifiable_from_orbit — stopped at visibility from orbit
213The reports locate it only to the country — too coarse to point a satellite at. — stopped at location

The same discipline runs through the Agent: see the full run where it refused to assess an event that never happened — factual claims carry their sources, and each brief reports its own citation coverage.

These documents describe how the product works; they are not a guarantee that any single report is true. The output is decision support, not confirmed intelligence. Read Source & Content Policy, Methodology and Trust & Security.

Pipeline activity — last 24 hours, as of 22:27 UTC

News read at scale, folded into events.

Articles are screened, and reports of the same event are folded into one source-linked event.

Articles read: 12,054
Outlets: 2,285
Events identified: 771
Countries: 85

This is what Delta read while no one was asking. A general AI assistant reads what it can fetch when you ask, and does not keep a running record of the events once the conversation ends. Why that matters, with the numbers

Discover · assess · observe · monitor · report

One intelligence loop. Multiple ways to use it.

Every surface answers the next question in the same loop — in the web app, over the API, or from your agent.

One signal layer underneath. World events (conflict, security, disasters, infrastructure) drawn from news worldwide and official hazard feeds, AI-enriched and refreshed every few minutes. The same signals feed the Watchfloor, the Agent and the daily report, and the Agent's findings open on the Map.

How these grew since January 2026

Developers, AI agents & AI assistants

Intelligence where your workflows already run.

Bring source-linked, geolocated events into dashboards, analytical pipelines, and AI agents through a REST API and an MCP server — the same intelligence the Watchfloor shows. Included on every plan; calls are metered in tokens (pricing).

Using ChatGPT or Claude? Delta does not replace the assistant: add it as a connector and the assistant answers from Delta’s record — what changed, where, and whether a satellite can check it — instead of re-reading the news. Connect your assistant

1 · FetchPull the events for a region or a category.
2 · CheckRead how precisely each one is located, and who reported it.
3 · WatchPut a place or an event under watch and read what changed.

python

import requests

headers = {"Authorization": "Bearer ond_YOUR_API_KEY"}
params = {}
resp = requests.get("https://offnadir-delta.com/api/v1/signals/5c2c299d-1aac-4fd8-9d3c-ff90b74bc121", headers=headers, params=params)
resp.raise_for_status()
print(resp.json())

bash

# Connect once — any MCP client (Claude Code shown):
claude mcp add --transport http off-nadir-delta \
  https://offnadir-delta.com/api/v1/mcp \
  --header "Authorization: Bearer ond_..."

# Then your agent calls the same intelligence as tools:
query_signals {"recency": "24h", "limit": 5}

json

{
  "id": "5c2c299d-1aac-4fd8-9d3c-ff90b74bc121",
  "occurred_at": "2026-11-07T00:00:00+00:00",
  "category": "infrastructure",
  "title": "Russian troops strike substations connecting Rivne and Khmelnytskyi nuclear plants to Kyiv",
  "location": "power transmission networks carrying electricity from nuclear power plants to Kyiv, the Kyiv region, and central Ukraine",
  "country_code": "UA",
  "lat": 50.4522,
  "lng": 30.4398,
  "plottable": true,
  "stage": "pinpointed",
  "stage_reason": "located_to_facility",
  "severity_score": 8.3,
  "geoint_score": 9,
  "escalation_trend": "escalating",
  "collection": {
    "level": "GSD<1m",
    "sensor": "vhr-optical",
    "analysis": "battle-damage-assessment",
    "niirs": "NIIRS 5",
    "target": "the struck site at Substations that connect the Rivne and Khmelnytskyi Nuclear Power Plants to Kyiv and central regions",
    "reason": "the struck site at Substations that connect the Rivne and Khmelnytskyi Nuclear Power Plants to Kyiv and central regions; vhr-optical + battle-damage-assessment to assess the strike damage.",
    "aoi": null,
    "gap": ".",
    "observability": 2.02,
    "observable": true,
    "observable_reason": "visible_at_observability_2"
  },
  "reporting": {
    "member_count": 4,
    "distinct_hosts": 4,
    "distinct_texts": 4,
    "source_reliability_best": "B",
    "credibility_best": 2,
    "graded_members": 4
  }
}

Get a free API key Open the Playground Connect via MCP

61 REST operations and 23 MCP tools available right now — counted from the live registry when this page rendered.

Satellite observations

What can available satellite data reveal?

Delta works with open satellite data: Sentinel-1 radar, Sentinel-2 optical, VIIRS nighttime lights and NISAR L-band radar (provisional calibration). It covers the globe on a regular schedule at roughly ten-metre detail. It is not high-resolution commercial imagery.

An assessment, not an observation

Every event states whether what was reported could be seen from orbit, and why not when it could not. That is a judgement made from the reporting. No image has been looked at to make it.

Imagery you open yourself

On the Map you open the scenes over the place and compare dates: radar through cloud and at night, optical in clear daylight. Sentinel-1 and Sentinel-2 are on the free plan; nighttime lights and NISAR are on paid plans.

Visible is not the same as true

An image can show that something changed on the ground. It does not show who did it or why, and seeing no change does not show that nothing happened.

Usually within reach

Flooding over a wide area, through cloud, with radar
Active fires and the burn scars they leave
Ships at a port, a strait or an anchorage
Large changes to the ground between two dates
A drop in nighttime lights over a city

Usually out of reach

Damage to a single small building
Anything that happens indoors
An event that leaves no lasting physical trace
The ground under cloud, for optical imagery

Open the Map Data sources Measured events

Use cases

What it is built for

Geopolitical & Supply-Chain Risk

Geolocated world events, AI intelligence briefs, daily reports, and satellite follow-up for risk, supply-chain, and insurance teams.

How it works

Defense & Maritime Domain Awareness

SAR ship detection, port and chokepoint monitoring, and maritime and security situational awareness with Sentinel-1 radar.

How it works

Disaster Response & Assessment

Flood mapping through cloud with SAR, live active-fire tracking, and before/after comparison of wide-area impact.

How it works

Developers & AI Agents

A REST API for event signals, briefs, and assessments — plus an MCP server that gives AI agents world-event awareness.

How it works

All use cases

Pricing

Start free — no credit card required. Usage is metered in tokens, so you pay only for the imagery and analysis you actually run. For teams, one shared token pool covers everyone on a single bill.

Explore & prototype free · monitor continuously · build & use commercially · operate as a team — the API & MCP server are included on every plan.

See plans & pricing Team plans

Free Tools & Guides

Live tools for checking an area yourself, right in your browser — no account needed — plus step-by-step guides for going deeper

Live Fire Map

WildfireLive tool — no login

Interactive NASA FIRMS map — view wildfire hotspots right on the page, dates back to 2000

Open tool

Free GeoTIFF Viewer

DataLive tool — no login

Drop a .tif and view it instantly — parsed 100% in your browser, never uploaded

Open tool

NDVI Vegetation Map

AgricultureLive tool — no login

Browse live global NDVI on the page, then calculate 10m Sentinel-2 NDVI for your fields

Open tool

Satellite Area Monitoring

Monitoring

Draw any area and track how it changes over time — no GIS, no code

Read guide

Change Detection

Analysis

Compare satellite images before and after any event

Read guide

SAR Flood Mapping

Disaster

Learn to detect and map floods using radar imagery

Read guide

SAR Ship Detection

Maritime

Identify vessels in SAR imagery for maritime monitoring

Read guide

Nighttime Lights

Urban

VIIRS Black Marble imagery for urban growth and power-outage analysis

Read guide

Getting Started Guide Use Cases FAQ

From the Blog

How to read geographic risk from satellite data — practical guides to SAR, spectral indices, change detection, and monitoring

Latest

remote sensingSAR

Can Satellites See at Night? Optical, Thermal and Radar

Can satellites see at night? Depends on the sensor: cameras need sunlight, VIIRS sees lights and moonlit clouds, thermal sees heat, radar sees in the dark.

Oct 29, 202610 min

Read article

resolutionsatellite imagery

Can Satellites See People, Faces or License Plates?

Can satellites see people, faces or license plates? 30 cm commercial pixels show a person as a dot and cannot read a plate. The physics, the specs, the limits.

Oct 29, 202610 min

Read article

remote sensingSAR

Can Satellites See Underground? Radar, InSAR and Gravity

Can satellites see underground? Only a little: radar reaches metres into dry sand, InSAR sees ground move above mines and tunnels, gravity sees groundwater.

Oct 29, 202611 min

Read article

Start here

Sentinel-1SAR

Getting Started with Sentinel-1 SAR Imagery: A Beginner's Guide

Learn how to access and analyze Sentinel-1 SAR radar data in your browser. This step-by-step guide covers the basics of SAR imagery, how to search for data, and interpret your first radar image.

Updated Oct 26, 20268 min

Read article

change detectionSAR

Change Detection from Space: Three Approaches That Actually Work

Not all change detection methods are equal. This post compares image differencing, NDVI time-series, and SAR coherence-based approaches with real-world guidance on when to use each one.

Updated Oct 18, 20267 min

Read article

monitoringtime series

How to Monitor Any Area with Satellite Time Series: A Complete Guide

Area monitoring with satellite time series lets you automatically track changes in any region over weeks, months, or years. Learn how to set up a monitoring zone, choose the right index, and interpret results using Sentinel-1, Sentinel-2, and VIIRS data.

Updated Jul 11, 202611 min

Read article

View All Articles

Frequently Asked Questions

What to know before you sign up.

A general AI assistant reads what it can fetch at the moment you ask, inside one context window, and does not keep a running record of the events once the conversation ends. Off-Nadir Delta reads continuously whether or not anyone asks: in the 24 hours to 19 October 2026 it took in 15,122 articles, about 45 million characters of text — more than ten 1M-token context windows. It folds reports of the same event into one record, keeps what changed with the source behind each change, places each event only as precisely as the reporting allows and says whether a satellite could check it, and records events that never reach the news, such as satellite-measured heat sources and changes in ship transits through straits. It does not replace your assistant: connect Delta over MCP or the REST API, and the assistant reasons over these records instead of re-reading the news.

View all FAQs

Start Reading the World Today

Jump back in — your Watchfloor, Delta Agent, the Map, and today's report are one click away.

Open the Watchfloor