Moderniser
Your AI agents are shipping code. Who's watching?
The governance layer for AI-driven software delivery — built for enterprises modernising SAP and other large systems.
The problem
Agents write code at machine speed. Approval still runs at meeting speed.
A developer with an agent can generate a week of changes before lunch. What breaks is not the code — it is the governance around it: nobody can see the plan, review it before it runs, or say what the work cost. Enterprises do not ship what they cannot audit.
Why Moderniser
Visibility, governance, control.
Visibility
Every agent session, plan and status change, on one board your whole company can read.
Governance
Two things cross the wire. Task text and wiki text. No code, ever.
Control
Metered, billed and attributed — down to the task.
How it works
The agent works where the code is. The plan lives where the reviewers are.
01
Your developer runs the Moderniser agent on their own machine, against checkouts they already have.
02
The agent plans the work and writes it to the board — tasks and subtasks, one per change.
03
A manager reads and corrects the plan in the browser. Corrections are attributed.
04
The agent executes; the ledger meters every call. Wiki pages capture what was learned.
Security
Security by architecture, not by promise.
We never hold your source. The strongest control is the one that cannot be bypassed.
Built for SAP
Decades of ABAP deserve better than a chat window.
Custom code, parallel ledgers, release windows measured in months — SAP estates are where governance matters most and is hardest to add. Moderniser's connectors speak to SAP Commerce, SAP ECC and S/4HANA with explicit read/write policies.
Governance
Six answers to the CIO's questions.
Plans on the record
The agent writes its plan to the board before it touches a task. You read it in the browser, not in a terminal scroll.
Corrections, attributed
A manager edits a task and the edit is theirs, timestamped. The board shows the last move of everything.
Company-wide visibility
Sessions, tasks and wiki pages are visible to everyone in the company by default. Creator is attribution, not access control.
Two things cross the wire
Task text and wiki text. No code, ever — the agent works on checkouts that never leave your network.
Connector policies
Every connector declares a policy: READ_ONLY, WRITES_WITH_APPROVAL or BLOCKED. Reads are safe until you say otherwise.
Money you can audit
Inference is metered at the gateway — the component that physically sits on every model call — and priced into a ledger per task.
Early proof
1
place to watch every agent
0
lines of code that leave
100%
of work attributed
Connectors
The agent doesn't just read your code. It reaches your systems.
SAP Commerce
OCC catalogue, products, builds
SAP ECC
ADT: packages, ABAP source, transports
S/4HANA
planned on the roadmap
Dynatrace
problems, entities, metrics
Salesforce
Sales Cloud and Marketing Cloud
Oracle ERP
Fusion Cloud + E-Business Suite
See it on your own landscape.
Twenty minutes, your systems, no code leaving your network.