Orbitthe platform by MAXQlabs
Do you really know what’s inside your software landscape?
Orbit reads your code and writes down what your systems really do. A living map of every application, with evidence to the line, on your own AI keys.
- < 1 weekfor an analysis that used to take months
- file:lineevidence behind every claim
- never stalerebuilt at every release
02The black box
Most organisations run on software nobody fully understands anymore.
The people who built it moved on. The documentation went stale the day after go-live. What remains is the code, the only place where the truth still lives.
- 01
Knowledge in a few heads
The people who understand the core are close to retirement. Onboarding takes years.
- 02
Documentation out of date
What exists describes an earlier version of the system, or a different one.
- 03
Dependencies unknown
Which programs touch which data, which jobs feed which reports: nobody has the full map.
- 04
Every change is a risk
Maintenance, compliance and modernisation decisions wait for an analysis nobody can finish.
COBOL is one example. Ten-year-old Java, .NET and data platforms have exactly the same problem.
03What it costs
Every decision waits for an analysis that takes months.
Every modernisation, migration or compliance question starts with an analysis phase nobody wants to pay for. So decisions get postponed, or taken blind: the migration that never starts, the audit that runs on assumptions.
Four forces make that blind spot expensive now.
- 01
Knowledge is retiring
The people who know why the system behaves as it does are leaving. The rules stay buried in the code.
- 02
Regulation wants proof
DORA, NIS2 and the AI Act ask for a current, provable map of systems, data and suppliers.
- 03
AI agents need ground truth
An agent can only act safely on an organisation it can describe. Most organisations cannot describe themselves.
- 04
Europe wants its own stack
Boards and governments are rethinking their dependency on platforms outside Europe.
04What Orbit is
Orbit reads the code and writes down what the system really does.
A team of AI agents ingests your source code and builds a typed model of each application: components, operations, data entities, workflows, integrations and findings.
- Evidence to the line. Every claim points to the file and line that prove it.
- Honest about certainty. Every claim is marked confirmed, inferred or an open gap.
- Never stale. Re-ingested at every release, so the picture moves with the code.
05How it works
From repository to evidence in five steps.
- 01
Connect
Git repositories or uploads, from Java and .NET to SAS and COBOL. On your own infrastructure.
- 02
Orient
A dossier per source: structure, subsystems, technologies. What a new architect would write in their first week.
- 03
Model
Components, entities, workflows and integrations, every claim with its evidence.
- 04
Assess
Checks such as NIS2 and GDPR run over the model. What fails becomes a concern, what passes an attestation.
- 05
Decide
Every finding bridges to a decision: keep, rewrite, migrate or retire.
Setup happens once. The first AS-IS of one technology lands within the week.
06See it at work
Orbit itself, reading a real data platform.
Four screens from a live Orbit environment. Documentation that takes months to write by hand, here written from the code, with the evidence one click away.
Landscape: what runs, who uses it, where it sits
- 1
One narrative per solution, written from the code: 22 components, 39 stores, 51 integrations.
- 2
Every claim carries its evidence marker and the source file it came from.
- 3
System overview, the structural map and the working assumptions, in that order.
Workflows: what the system does, end to end
- 1
31 workflows reconstructed from the code, grouped per business domain.
- 2
Each one names the components it touches and carries its own evidence status.
- 3
The processes that lived in people’s heads, now written down.
Entities: what the solution stores, and how it links
- 1
49 entities with their relationships, per domain and per component.
- 2
Fields carry their sensitivity: where every GDPR question starts.
- 3
Flow and organic views of the same data model.
Components: from overview to the exact line of code
- 1
22 components, each with its kind, domain and open findings.
- 2
Runtimes past end of life are flagged automatically.
- 3
From a component to the finding, the file and the line number.
07Honest by design
Every claim shows where it came from, and how sure we are.
Orbit never asks you to trust a description. Each statement points to the code that proves it, and says how certain it is. Where the code cannot tell, Orbit says so, and asks a person.
- confirmed
- Verified directly against the source.
- inferred
- A plausible reading. The reasoning is written down, so anyone can check it.
- gap
- Not determinable from the code. It becomes an open question for a person.
Nothing is believed because it is written down. When the code changes, Orbit looks again.
-
Invoices above €10,000 need a second approver before payment.
confirmedfile:billing-api/src/Approval.cs:45-60 -
Rejected invoices are archived after 30 days.
inferredfile:jobs/archive_rejected.sql:12reasoning recorded -
Who may override a rejection?
gapopen question → finance team
08The proof
Analyses estimated at 3 to 12 months landed in under a week.
Classic analysis estimate With Orbit
09Faster, and better
The same method every time gives a more consistent and more complete result than manual work.
Traditional manual assessment
- Months of interviews and reading by scarce experts
- Quality depends on who looks, and where
- Out of date the day it is delivered
- Rarely repeated, because it costs too much
AI-first, with Orbit
- Days to weeks, on the code itself, with the same method for every application
- Every finding with evidence to the line
- Rebuilt at every re-ingestion, never stale
- Repeatable: a foundation that stays current after the first report
You stop buying effort. You start buying a result with predictable quality.
10Built for legacy
Including the code nobody wants to touch.
Orbit reads modern stacks and mainframe estates alike. On a COBOL environment, programs, copybooks, JCL and data definitions become an AS-IS per program, on your own infrastructure.
- COBOL
- JCL & copybooks
- SAS
- Java
- .NET
- SQL & data platforms
- AS-IS
Per program, in plain language
Business logic, inputs and outputs, with evidence to the line.
- MAP
Dependencies and flows
Call graphs, data lineage and batch chains: the map nobody had.
- FINDINGS
Risks you can point at
Dead code, risks and undocumented rules, each traceable to file and line.
11From insight to action
One foundation. Three moves.
Insight is only useful if it leads somewhere. The same AS-IS carries all three moves, with no new ingestion in between. And where there is nothing to analyse yet, Orbit builds new software straight from intent: spec, tests and code. Nothing merges without green tests and a person signing off.
- 01
Analyse
An AS-IS of the whole landscape: what runs, what it does, what depends on what.
- 02
Evolve
AI-assisted maintenance on that foundation. Every change starts from what the system really does.
- 03
Migrate
The AS-IS as your migration base: you know what can go, and what must come along.
One landing, then growth
- ↑ UpFrom one application to landscape insights and management views.
- ↓ DeepInto the AI-assisted lifecycle, per unit that keeps changing.
- → AcrossThe next technology, on the same platform.
Compliance (NIS2, GDPR, DORA) runs across all of it, on the same data.
12Your data stays with you
Orbit runs on your keys and your infrastructure.
-
Bring your own key
AI consumption runs through your own model provider, on your own keys.
-
The code never leaves the house
Orbit runs in your environment, in the EU. Your code is never used to train models.
-
Your identity provider
Single sign-on through the directory you already run.
-
Every step on the record
The agents commit every step they take: dated, reviewable and reversible.
-
Code, not data
Orbit works from code and specifications. Customer and production data are never an input, so no personal data goes to the AI.
-
Yours to take, no lock-in
What Orbit delivers, the analysis and the code, belongs to you. Plain production code, with no AI running inside your applications.
- Built to EU AI Act, GDPR and ISO 27001 standards
13Next · Orbit Landscape
From one application to the whole estate.
Organisations don’t run one application. They run hundreds. Orbit Landscape joins them into one living map, observed from the systems themselves.
Landscape tools today
Declared
Fact sheets filled in through surveys and workshops. Out of date the day after.
Orbit Landscape
Observed
Systems, links and data flows read from the code itself, with evidence, refreshed at every release.
Change one system. See who feels it.
Every link knows how critical it is and what happens when the far end fails. So an impact analysis tells you what stops, what slows down and what merely notices.
Pick a system to change
Illustration on a fictional estate.
A change to ERP reaches 7 of the 9 other systems: WMS stops; Webshop, Finance and Payroll are delayed; Customer portal and Reporting degrade; the Data platform notices.
-
Systems and typed links
Every dependency and integration becomes a link, with its criticality and failure mode.
-
Impact across systems
Change one component and see which systems, capabilities and processes feel it.
-
Questions that stay answered
Ask once, in plain language. Every refresh stores a dated answer you can cite, and compare with last quarter.
-
Works with what you have
Feeds the enterprise-architecture tools you already use, with evidence instead of surveys.
One map, four obligations.
Regulators already ask for this map. Today it is assembled by hand, once a year. The landscape keeps it current, with evidence.
- DORARegister of information: every ICT service mapped to the business functions it supports.
- NIS2Asset management: a current inventory of systems and their dependencies.
- GDPRRecords of processing: which personal data lives where, per store and per flow.
- EU AI ActInventory: where automated decisions run, and what data they use.
14Where this goes
Start with one application and grow to the operating picture of your organisation.
-
Now
Orbit
Per application
A living AS-IS from the code, with compliance assessments.
- Confirmed, inferred or gap on every claim
- Evidence to the line
-
Next
Orbit Landscape
The whole estate
Systems and links across the portfolio, with impact analysis and dated answers.
- DORA, NIS2 and AI Act inventories
- Feeds your EA tooling
-
Later
Orbit Ontology
The operating picture
Business objects, actions and agents on top of the landscape graph.
- Derived from the landscape graph
- Sovereign and European
No big bang. Each horizon is used on its own, and the next one reuses everything already mapped. You never pay twice for the same knowledge.
Four layers, each one valuable on its own.
Read from the bottom up: each layer is built from the one below, so the top never floats free of what really runs.
Your people and your AI agents
- Later · action
Actions and agents
Governed actions and AI agents that work through the ontology, updating the systems themselves, with a full audit trail.
- Later · meaning
Business ontology
Customer, Policy, Claim: linked to the code that implements them. Proposed by AI, approved by people.
- Next
Landscape graph
Every system and link across the estate, with impact analysis and dated answers.
- Today · foundation
Code graph
Components, entities, workflows and findings, with file-and-line evidence.
15Later · Orbit Ontology
A working model of the business, read from the systems themselves.
An ontology holds the objects your business talks about, the links between them, and the actions people and AI agents may safely take. Orbit already maps where each object really lives. The next step lifts that map to the language of the business.
Business objects
Mapped by Orbit
CUST_MASTERTBL_POLCYCLM_MAINAI answers better when it reasons over structure.
More than three times better, from structure alone. The ontology is what makes AI safe to act on.
Sequeda et al., data.world benchmark, 2023 (arXiv 2311.07509).
Derived from what already runs.
Ontologies are usually modelled by hand, in weeks of workshops, and start ageing the day they are finished. Orbit starts from the other end: the landscape graph already knows which systems hold the customer, where the claim table is and which code changes it. It proposes the ontology; people approve it. European, and on your own keys.
Insurance
Policy · Claim · Customer
Trace a claim end to end. Show regulators where pricing rules live.
Public sector
Citizen · Case · Benefit
Once-only data across agencies. See which code implements which article of law.
Retail
Product · Price · Order
One price truth across ERP, tills and web shop. Order agents that act in the systems themselves.
Manufacturing
Order · Batch · Asset
Trace a quality incident to the code and configuration behind it.
16Questions we hear
The questions we hear most, and our answers.
We already have an enterprise-architecture tool.
Keep it. Orbit feeds it evidence from the code instead of surveys and workshops.
Large language models hallucinate. Why trust this?
Because you don’t have to. Every claim carries file-and-line evidence and a status. What the code cannot prove is marked as a gap and turned into a question for a person. Interpretations are proposed, and people approve them.
Is our code safe?
Orbit runs on your own AI keys, in the EU, with single sign-on through your identity provider. Your code never leaves your environment and is never used to train models.
How fast do we see results?
Setup happens once. The first AS-IS of one technology lands within the week. For a single application, a 20-day analysis becomes one day.
Which technologies does Orbit read?
From Java and .NET to SAS and COBOL, including data platforms. You start with one technology and grow to the next on the same platform, without ingesting anything twice.
Does it replace our architects and analysts?
No. It gives them the ground truth in days instead of months, so they spend their time on decisions.
17Your landscape scan
Do you really know what’s inside your software landscape? Let’s find out, on your own code.
One technology. Your code, your keys. The first AS-IS within the week.