Every instrument in the catalogue arrives in one of three shapes, and the shape decides what “install” means. This page describes the intended mechanics for each. Nothing has shipped, so read the steps below as the specification being built to — not as a walkthrough you can follow today.
The three delivery shapes
On-platform instruments run natively inside the trading environment you already use. They are sold as standalone marketplace SKUs, and the marketplace handles delivery, licensing and updates the way it always has. Installation is that platform’s own flow: acquire the SKU, place the file where the platform keeps its add-ons, restart it, and the instrument appears in the platform’s own list. No Hadal account is involved in that path, and the measurement runs against data already on your machine.
Hub instruments run in a browser. There is nothing to download and no version to keep current. You sign in, point the instrument at a file or a connection, and the run happens server-side. The Overfit Auditor, the Epistemic Harness and the Agent-Governance Constitution are hub-only, because the work they do has no natural home inside a charting platform.
Integrity services are not installed at all. The Dashboard Lineage Audit and the Gate Falsification Kit are engagements: you hand over a system and its claims, and what comes back is a report against a published method. There is no binary in that transaction.
Which shape an instrument takes is stated on its own page. The suites index groups all twenty-one by family if you want to see the split at a glance.
Installing, step by step
The sequence below is the same whichever shape you are dealing with. Only the second step changes, and it changes completely — which is why the first step is identifying the shape rather than downloading anything.
Identify the delivery shape
Open the instrument’s own page and read its delivery line. On-platform instruments install through your trading platform’s add-on flow; hub instruments have nothing to install at all; integrity services are engagements with no binary in the transaction. Every later step depends on which of the three you are holding, so this is not a formality.
Acquire it through the channel that shape uses
An on-platform SKU is acquired from the marketplace that lists it, and that marketplace handles delivery, licensing and updates exactly as it does for any other add-on — no Hadal account is created and no credential changes hands. A hub instrument is reached by signing in, so there is no local version to keep current and no update to miss. An integrity service begins with an agreed scope rather than a download.
Prepare the sample you intend to hand it
Assemble the window you want measured and be able to state its boundary: symbol, venue, dates, session hours. Use the raw capture rather than an aggregated export, and write down what you already changed about the strategy or the data before exporting it. The full preparation list is the next section, and it is the part of the install that actually determines whether the run is worth anything.
Prepare somewhere to keep the artifact
Every run is intended to produce a file carrying the inputs, the method version and a content hash over the whole thing. Decide where that file will live before you generate the first one. A result you cannot find again is a result you cannot defend, and the archive is what makes recomputation possible months later.
Read the limits before the first run
Each instrument page carries an explicit statement of what its output does not establish, and each battery declares a completeness bar below which it will refuse to estimate. Read both before the first run rather than after the first surprising result — see what a result does not establish and when an instrument refuses to answer.
What to have ready
The install itself is the short part. The measurement is only as good as what you feed it, so the preparation matters more than the download:
- A data window you can describe. Which symbol, which venue, which dates, which session hours. An instrument that cannot state the boundary of its own sample cannot publish an honest result, and it will ask you for that boundary rather than guess.
- The raw file, not a summary. Aggregated bars discard the structure most of the feed and execution batteries are looking for. Where an instrument needs ticks, a daily close will not substitute, and it will say so instead of running anyway.
- A note of what you already changed. If you re-fitted a strategy nineteen times before exporting it, the backtest-honesty instruments need to know. That count is an input, not an embarrassment — see reading a result.
- Somewhere to keep the output. Every run is intended to produce an artifact you can archive and re-check later. Treat it like a lab notebook entry, not a screenshot.
What installing does not do
Installing an instrument does not connect it to your broker, your account or your funds. Nothing in the catalogue places orders, manages positions, or holds credentials that could. The on-platform SKUs read what the platform already shows them; the hub instruments read what you upload. There is no path from a Hadal measurement to a trade except the one you walk yourself, deliberately.
It also does not entitle you to a result. An instrument that receives a sample too thin to support its estimator will refuse to produce a number — that behaviour is described in when an instrument refuses to answer, and it is the point rather than a defect.
Current status
There is nothing to install today. Checkout is not open, no SKU has been listed, and no download exists. The pricing model explains the structure that will apply when it does, including the rule that the hub price for an instrument sits at or below its marketplace-equivalent. If you want the install instructions when they become real rather than intended, the launch list on that page is where they will be announced first.
Documentation — page 1 of 9, in reading order