A scene can read the systems your customers run
By default a scene reads what is in your own workspace. It can also read what is in theirs: the S/4HANA instance behind their firewall, the ticket queue your desk works in, the historian on their plant network.
Most of it is a connection your customer authorises once, the same way they connect anything else. For the systems with no door out, a node goes inside their network instead. Either way you are not the one holding their credentials, and the answer comes back to the screen your engineer already had open.
The steps a run of this kind takes. No timings or findings are shown here, because the ones worth printing are the ones measured on a real system.
Three companies describe the same afternoon
Each one has people who spend it reading a system that belongs to somebody else.
The second-level engineer
Logs into the customer's instance, finds the document, compares it against the config that was supposed to govern it, checks whether the field they added is the thing that broke it, writes it up. An hour, per ticket, and the next customer's instance is different.
The consultant on day three
Still mapping the landscape by hand: which tables were extended, what the interfaces do, where the exceptions live. The engagement was sold on the advice that comes after this part.
The service desk for the machine
The line stops, the telemetry that would have shown it coming sits on the customer's own network, and the work order gets written from a phone call instead.
Two ways in, and the install is the rarer one
Connect it. An instance URL and a service key for their S/4HANA, a communication user, or the ticket system's own consent screen. Your customer authorises it once and it is an asset in the scene, the same shape as every other integration we ship. Nothing is installed and nothing is opened. This is how most of them start and how most of them stay.
Or put a node inside. For the systems with no door out: the ERP behind the firewall, the historian on the plant network, the database with no public endpoint. One binary, 45 MB, no runtime and no toolchain on the machine it runs on. It keeps a read-only copy of what it may see and dials outward, so nothing on their side is opened towards it, and the read path is the ABAP data-preview endpoint, so there is no database login in the chain to hand over, lose or rotate. We have run one in a production tenant since August.
Reading the system is not the slow part any more
Three numbers off live systems, each measured once. We would rather show you the run than the number.
12.4 million rows of it, read end to end. The engineer looking for the one row that matters opens four transactions instead.
112 GB of them, matched back to the records they belong to: 98.7% of purchase orders, 99.9% of requisitions.
Surfaced by one rewritten statement. The source endpoint caps statements at about 255 characters and has no window functions; the local copy has both.
Your keys, your data, your name on the screen
Your model contracts are tried first and ours sit underneath at list price if you want the fallback. The data and the traces are yours, and we do not train on them. Writes stop and ask, the way step 05 above did. Every run is recorded and reviewable, including the ones that got nowhere, so a wrong answer is something you open rather than a score you have to trust.
There is no price on this page
Nobody is buying a subscription here, so there is no subscription revenue for us to take a share of and the self-serve terms we publish are not the ones that fit. What this costs is built from the specifics: how many people it serves, whether the node sits in one network or three hundred, and what your agreement has to say.
What gets asked in the security review
Do you need access to our customers' systems?
Not from you. The customer authorises the connection themselves, or installs the node if the system has no door out, and either way they decide what goes on the list. You never hold their credentials.
What about customers who will not connect anything?
Then this does not help them. There is no version of reading a system that works without a way in, and we would rather say so here than in month three.
Does this replace our ticketing?
No. It feeds what you already run. The incident arrives where it always did and the note goes back to the same place.
Who is the data processor?
Where we operate the servers, you sign a data-processing agreement with us and your customers sign theirs with you. On your own servers the data never reaches us.
Does our name or yours appear?
Yours. Your sign-in, your domain, your screens. Which consent screens carry which name is part of the setup rather than something you inherit.
Can we run it on our own servers?
Yes. It is the same instance configuration on your machines: you take the server, the database and the vault.
Bring a ticket nobody could close
A real one, from a real customer instance. That is a better first meeting than a demo of ours.