Kortex
snapshot · 2026-09-06

Know which sites deserve a connection study.

The facts are public. The connections are the product.

Bring a shortlist of data-centre, generation or storage sites and Kortex screens every candidate against grid evidence, EHV topology, interconnection queues, operator-published capacity, water constraints and counterparty ownership; each site receives a defendable verdict, with every material finding sourced and dated: one sourced query, not weeks of joining datasets.

On the first call we run it live against the graph; you leave with the screened shortlist and the evidence behind every row. Prefer to try it yourself first? A free evaluation API key is issued instantly: 14 days, no call required, and it includes a private trial workspace for saved shortlists, evidence and daily monitors.

248 named sources · 15 published validation studies, misses included · every answer frozen to a retrievable receipt

One query, one traversal6 domains
? Which zones deserve a 300 MW connection study?
  1. scan →grid zonewholesale price, fuel mix, installed dispatchable low-carbon MW
  2. nearest_substation →substationinterconnection point & EHV topology
  3. located_in ←generatorsdispatchable low-carbon MW within reach
  4. operated_by → parent →counterpartyresolved owner, group concentration
  5. overlay →queue + watercongestion & completion history (zone-grain queue history US-only today; zones without it say so; the candidate-site screen also joins the GB, Irish and German registers at radius grain), stress
  6. answerzones ranked, with the evidence attached src EIA · ENTSO-E · GLEIF · queuesfreshness per sourceconfidence tiered

Illustrative traversal; every hop is a real edge type; run it live from /docs.

the proof · a real screen, re-run as the graph grew

Two of five site verdicts changed when the public connection registers were joined.

The same five candidate sites were screened three times on 12 July 2026 as new public registers entered the graph, each run frozen by its own receipt. The Leeds candidate flipped proceed → investigate when the GB Embedded Capacity Register exposed 11.2 GW already accepted-to-connect within 50 km. The Dublin candidate flipped when the operator's own published figures exposed 0.0 MVA demand availability at the nearest connection-grade point.

Every one of those facts was public all along. None of them was decisive until it was connected to the site. See the full screen, every verdict and its evidence; or compare the frozen runs yourself.

run 1 56da6c4arun 2 19e2acc7retrieval shared links (customer receipts are private by default)

the product · that re-run is now a standing capability

Screen once. Keep the shortlist live.

Save a shortlist as an investigation and Kortex re-screens it as public evidence changes: on demand, or on a daily, weekly or monthly monitor. Every run is frozen by its own receipt and chained to the previous one; when a verdict moves, the changes feed names the site, the check that drove it, and the evidence that appeared or vanished. Previous runs are immutable; editing a shortlist never rewrites history. Notes, confirmations and challenges attach to any finding; they stay private to your organisation and never alter a verdict.

The proof above is exactly this workflow: the Leeds candidate flipped proceed → investigate when newly joined register evidence exposed 11.2 GW accepted-to-connect within 50 km, with the old and new receipts linked. Open the workspace; access is provisioned per organisation via the pilot form.

example flip Leeds · proceed → investigatedriving check connection queuewas 56da6c4anow 19e2acc7

Kortex is an infrastructure-intelligence platform; siting large grid connections is its first application. Underneath sits the world's power system as one queryable graph: assets, grid capacity, prices, interconnection queues, ownership, water risk and geopolitical exposure, connected and source-traceable. Use it to screen sites, counterparties, portfolios and grid constraints in minutes instead of weeks.

Kortex is built by Orkora, a funded floating-nuclear developer, to screen sites for its own gigawatt-scale plants; the same evidence layer is what we sell.

How a siting pilot works

Siting a large grid connection (a data centre or large load, a generation project, a storage asset) is the one decision that needs every domain Kortex connects, before development capital is committed.

01 · you bringA shortlist10–20 candidate sites, or a region you're screening.
02 · Kortex traversesSix domains at onceInstalled dispatchable low-carbon capacity × queue congestion × wholesale price × EHV topology × water stress × counterparty ownership.
03 · you getA screened shortlistEvery site gets an evidence-based verdict (proceed, investigate, reject or insufficient evidence), every material finding traceable to a named source and dated; a receipt you can forward.
04 · the decisionConfidence before spendRanking confirmed, ranking changed, or weak sites killed before the money goes in.

Same pattern re-points at counterparty and portfolio diligence; bring the entities or assets instead of the sites. Siting is the front door because it needs all of it at once.

Auditable, reproducible, source-traceable. The answer path is deterministic: the same query against the same graph returns the same answer, traceable to named source datasets; no LLM, no generative AI, no black-box scoring in the answer path. Where a value is modelled rather than observed, the response says so; every material metric carries an evidence class.
Screening, not forecasting. Kortex surfaces current-state compound risk, competitive dynamics and structural exposure from historical and real-time data. No forward price curves, no demand forecasts, no scenario models; it complements forecast-oriented research platforms rather than replacing them.

The platform, in one block

Applications. Site screening for large grid connections; load, generation and storage (live) · infrastructure due diligence & portfolio monitoring (delivered through pilots, productising).

Shared intelligence substrate. Assets · ownership · grids · markets · water · hazards · countries and macro series, joined through one typed graph.

Five governed data surfaces. Tables · time series · country indicators · asset profiles · entity profiles: the whole servable estate, rights resolved live on every response, commercial accounts failing closed on non-cleared sources by default.

Evidence. Coverage · methodology · 15 validation studies (misses included) · rights modes · retrievable answer receipts · correction history.

The API. Typed, self-describing REST in four layers (insights, graph, time series, reference) with an MCP server for AI agents and per-query cost in the response; full reference at /docs.

The numbers behind this block (nodes, edges, datasets, a real record up close) live on the platform data page; revenue modelling methodology at /methodology, one register for how every derived number is made.

Who it is for

UserWhat they use it for
Siting & power-origination teams (load, generation, storage)Grid zones ranked by installed dispatchable low-carbon capacity, queue congestion, EHV connectivity, water stress and counterparty exposure; before committing development capital.
Infrastructure funds & lendersPre-diligence screening of assets and counterparties: transition exposure, stranded-asset risk, physical risk, ownership-chain concentration.
Grid & energy consultanciesCross-engagement data infrastructure: market entry, queue assessment, retirement screening; without rebuilding the joins per project.

The same graph also serves development finance, ESG and transition-risk analysis, commodity desks, and AI agents via the MCP surface; ask us about your workflow.

Quick start

Explore.

Global Viewer Platform data Sources & licences API docs

Try live, no key. Rate-limited, exact queries only; full parameter access comes with an API key.

Queue completion Fossil revenue exposure Top grid zones

The next step

Bring 10 to 20 candidate sites, or the one standing question your team already asks every week. On the first call we run it live against the graph; you leave with the screened shortlist and the evidence behind every row. It succeeds if it changes a decision, finds a material fact, or materially compresses the work.

Screen a shortlist How pilots work

Contact

Orkora Ltd · contact form · kortex@orkora.com · free 14-day evaluation keys issued instantly · production access by agreement

The fastest evaluation is a technical demo on assets or regions you already know well: we run Kortex queries live and you validate the outputs against your own data.