Orinova
Service
Turn legacy
code into a living knowledge asset.
Orinova reads your mainframe COBOL — and your VB6 and classic .NET estates — then answers your team’s questions with source-cited precision. AI drafts the documentation; your engineers confirm it; from confirmed docs, Orinova derives a Java target design with full traceability. In DEHA-delivered projects, this has cut migration cost and duration by 30–50%* and estate visualization from months to days.
●
●
●
If we change the customer-code field in the closing batch, which reports are affected?
Any dead logic in that module?
Four reasons modernization stalls.
Orinova unblocks each one.
01
“The engineers who understood this system have retired.”
Orinova rebuilds their knowledge from the code itself – specs, business rules and dependencies reconstructed by AI, kept in a knowledge base that stays with your team.
02
“No design docs. Nobody dares touch anything.”
Auto-generated specifications, flow diagrams and impact analysis replace fear with evidence – every answer cited back to the exact line of code.
03
“Assessment alone takes six months of expert time.”
Bulk-ingest the whole estate and visualize its structure in days. Dead-logic detection trims migration scope before conversion even starts.
04
“We can’t put our source code in someone’s cloud.”
Deploy on-premise or in your private cloud, with fully local LLMs if needed. Your code is never used to train external models.
Built for real modernization projects.
30–50
%
Lower cost & duration*
Days
not months
Estate visualization
100
%
Answers with citations
2x
Throughput in one re-platform case*
Understand. Document. De-risk.
Load the latest source and surviving documentation – Orinova parses, indexes and analyzes everything, then lets your team interrogate the system in natural language.
CAPABILITY 01
Ingest the whole estate
Upload as zip or connect Git/SVN directly, with incremental sync. For COBOL, the Codebase Profiler auto-detects dialect, CICS, DB2, VSAM and more, and builds the CALL/COPY/PERFORM dependency graph. PL/I, VB6, classic ASP/.NET, Java, design docs, spreadsheets and PDFs (built-in OCR) are ingested as searchable, citable knowledge.
CAPABILITY 02
Documentation your engineers sign off
Per-program docs – purpose, data flows, business rules – cited to exact source lines, each with an AI confidence score. Engineers review, edit and confirm every document, prioritized by program criticality, with versioning, diff and rollback. AI drafts; humans decide.
CAPABILITY 03
Impact analysis, on demand
“If I change this field, what breaks?” — answered with affected programs, jobs, screens and data stores. Grounded, cited, auditable.
CAPABILITY 04
Dead logic, detected
Unreachable and obsolete code paths identified before conversion begins. Shrink migration scope – and the bill – with evidence.
CAPABILITY 05
Ask the system anything
Specs, modification approaches and business rules in plain language, in six UI languages. Every answer carries inline citations, grounded in your registered assets and checkable against source.
CAPABILITY 06
Enterprise-grade, day one
Multi-workspace with schema-level data isolation, RBAC, SSO, enforceable TOTP 2FA, audit logs, per-workspace AI cost governance, API & webhooks. Local LLMs via Ollama – or your own OpenAI-compatible endpoint – for air-gapped sites.
Ingest to Java design.
Code generation, when it’s earned.
We’d rather tell you exactly what’s ready than blur the line. The knowledge platform, COBOL analysis and documentation, and Java design with traceability are shipping today. Code generation is our next phase – and it will start from your engineers’ confirmed documentation, not from raw source.
STAGE 1
Ingest
Zip upload or direct Git/SVN connection with incremental sync.
STAGE 2
Analyze
Codebase Profiler auto-detects dialect and environment; builds the CALL/COPY/PERFORM dependency graph.
STAGE 3
Document
Line-cited drafts with confidence scores – reviewed, edited and confirmed by your engineers.
STAGE 4
Java design
Controllers, services, entities, batch – plus a COBOL→Java rule-mapping matrix with full traceability.
STAGE 5
Code generation
Not available today, by design. Generated code is only as good as the verified understanding beneath it – so we built the understanding first.
One platform, five steps to modern.
Orinova powers every phase of DEHA’s modernization methodology – from first assessment to final handover.
STEP 1
Assess & evaluate
Visualize the full estate in days. Surface risk, estimate ROI.
STEP 2
Design the target
Accurate baseline for microservice & cloud-native architecture.
STEP 3
Prove it (PoC)
Pick safe, representative functions. Verify boundaries first.
STEP 4
Migrate & integrate
Engineers query Orinova daily; specs stay current as code moves.
STEP 5
Hand over, alive
A living knowledge base – not a PDF snapshot – enables self-run teams.
Orinova vs. other
paths to modernization.
There is more than one way to leave a legacy system behind: automated converters, large integrators, hyperscaler toolchains, or teams of consultants reading code by hand. Each has its place, and we won’t pretend otherwise. What sets Orinova apart is the starting point: your source code stays inside your own network, on your own terms, while Orinova builds a verifiable, line-cited model of what your system actually does – before anything is converted or rewritten. Migrations don’t fail at conversion speed. They fail on incomplete understanding.
“Can’t we just run a converter and get modern code?”
If the goal is syntactically working code, fast, converters deliver. Orinova is not a converter: it produces a Java target design and a traceable COBOL-to-Java mapping matrix from documentation your engineers have confirmed.
Converters translate one-to-one – dead code, workarounds and duplication included. Orinova answers the question they skip: what should be migrated, retired or redesigned, every finding cited to source. Its outputs feed whichever converter or integrator you choose.
Before you convert a single line,
know which lines deserve to survive.
“Shouldn’t we hand this to a large integrator end to end?”
For delivery scale and mission-critical track record, the major integrators are genuinely hard to match.
But their assessment arrives as documents, priced by the person-month, and walks out when the engagement ends. Orinova installs the capability inside your walls: a permanent, queryable model your team owns — neutral on where you migrate and who executes it.
Keep the understanding in-house -
whoever you hire to do the migration.
“Haven’t the hyperscalers already built this?”
If your destination is already fixed to one cloud, their tooling is deep, well funded, and worth evaluating.
Those tools assume your code can leave your premises and your destination is their platform. Orinova assumes neither: fully air-gapped if needed, the LLM of your choice, neutral on target architecture – with a delivery team included.
Deep code analysis that doesn’t require
handing over your source code — or your roadmap.
“Doesn’t this really take experienced engineers reading the code?”
It does – and it always will. Expert judgment can’t be automated.
What doesn’t scale is asking humans to read millions of lines. Orinova scans the whole codebase, extracts rules with line-level citations and turns your experts from authors into reviewers – role-gated confirmation, confidence scores and an audit trail of who approved what.
Let your experts verify instead of excavate.
Not ready to commit to a migration?
You don’t have to be. The first phase is read-only analysis inside your perimeter – fixed-scope, fixed-price – and you keep current-state documentation whether you migrate next year or not. After cutover the same deployment continues as your enterprise AI knowledge base on the Oriene platform.
Built to survive a bank’s due diligence.
One platform core.
A constellation of products and services.
Oriene AI
Enterprise AI chatbot (RAG) grounded in your own knowledge base.
Orinova
AI-powered legacy analysis & modernization. In Japan: BIZモダナイズAI.
Orion CRM
AI-first CRM, self-hosted.
Oriflow Ops
Autonomous multi-step agents for operations.
Oriac Agents
Domain AI agents by industry.
Oridex
Turn scattered documents into structured, searchable knowledge.
Questions, answered.
Can we run Orinova on-premise or air-gapped?
Yes. Deploy on-premise or in your chosen cloud region. With local LLMs via Ollama, Orinova runs in fully internet-disconnected environments – nothing leaves your control. You can also point Orinova at your own OpenAI-compatible endpoint.Yes. Deploy on-premise or in your chosen cloud region. With local LLMs via Ollama, Orinova runs in fully internet-disconnected environments – nothing leaves your control. You can also point Orinova at your own OpenAI-compatible endpoint.
Which languages and formats can it ingest?
The deepest pipeline – Codebase Profiler, dependency graph, Java design and mapping matrix – targets COBOL. PL/I, VB6, classic ASP/.NET, Java and document assets (spreadsheets, PDFs, scans via OCR) are ingested as searchable, citable knowledge for documentation and Q&A. Ask us about coverage for your specific stack.
How do we know the answers aren’t hallucinated?
Every answer carries inline citations to the exact code line or document passage it came from. A knowledge-base-required mode restricts answers to your registered assets only.
Does Orinova generate Java code?
Not today. Orinova produces a Java target design and a traceable COBOL→Java mapping matrix from engineer-confirmed documentation. Automated code generation is on our roadmap, with no announced date – we ship it when it can start from verified understanding.
How do we know an AI-generated document is correct?
Every draft carries line-level citations and a confidence score used to prioritize review. Engineers confirm each document through a role-gated workflow (RBAC), with versioning, diff and rollback, and an audit trail of who approved what. The confirmation step – not the score – is the guarantee.
Do we have to buy a migration project too?
No. Orinova works standalone for knowledge recovery and de-risking. It’s most powerful combined with a DEHA-led migration to Java, C#, Node.js or Python on AWS or Azure – but that’s your call.
Is our source code used to train AI models?
Never. Your code and data are not used to train external models. Delivery teams operate under ISO 27001 in dedicated lab environments.
How fast can we see it on our own code?
We load one representative COBOL subsystem, run the Codebase Profiler and dependency analysis, and evaluate the results together with your engineers – typically 2–4 weeks.
See Orinova on your legacy code.
A 2–4 week proof of concept: we load one representative COBOL subsystem, run the Profiler and dependency analysis, and evaluate the results together.