Derive the right
architecture.
Human engineering judgment and AI assistance on one connected model, from concept architecture to a connected digital twin, with deterministic evidence behind every decision.
- 14 daysfree trial, every feature
- Ask · Plan · BuildAI copilot modes
- SysML v2 · ReqIFopen-standard exchange
- Windows 10 & 11installs locally, database included
For architects, system designers and MBSE practitioners. Windows 10 and 11, no payment details, your models stay on your machine.
The harness
One harness around
the whole MBSE flow.
Human judgment and AI assistance accompany every step, from the first concept to the connected digital twin, not one tool bolted on at the end. Built for automotive, aerospace, heavy equipment and medical systems.
Concept
Frame the problem and the candidate concepts.
System intent & requirements
What the system must do, traced from the start.
Logical architecture
Functions, ports and the items they exchange.
Physical architecture & networks
ECUs, wires, buses and the signals they carry.
Evidence
Checks, bus load and trade studies against your limits.
Digital twin
A connected model that stays the reference.
The harness · at every step
Human judgmentYou decide, review and sign off.
AI assistanceAsk · Plan · Build, with every change reviewable.
What changes
Derive the decision,
with the evidence attached.
Architecture work is revisited for years. Keep the reasoning in the model, not in slide decks and spreadsheets that drift apart.
-
01 Architecture
Derive the architecture, from intent to wire.
Start from system intent, derive the interfaces, then realise them on ECUs, connectors, wires and networks. Each step stays traced to the one before it.
-
02 Evidence
Derive the evidence from the model itself.
Checks, bus load, voltage drop and trace coverage are computed from the model, not typed into a report. A check that tested nothing says "not evaluated", never "passed".
-
03 Results
Derive deterministic results you can repeat.
Same model, same inputs, same answer. Solver runs are recorded with their seed, the change history is hash-chained, and a restore point is captured before the AI writes.
AI assistance, with you in charge
AI proposes. Deterministic engines prove.
You decide.
D3rive exposes the verbs an engineer uses (query the model, check it, solve an allocation, compare candidates) as tools an AI can call. The AI drafts; the checks, computations and solver runs behind its proposals are deterministic and recorded; you review, compare and decide.
- Every AI action runs with your permissions, your quotas and the same validation as your own edits
- Answers cite the model elements they were drawn from
- Changes arrive as a reviewable plan, and Build records a baseline you can restore first
- The copilot never deletes, and every change it makes is attributed in the history
- Ask
Answers drawn from the model, with links to the elements behind them.
- Plan
A reviewable change set you inspect before anything is written.
- Build
Applies the plan after capturing a project baseline you can restore.
- Your own agent
Drive the same tools over MCP with a personal access token.
Backers wanted
Join Early Access
To everyone who already backs us: thank you.
Become a backer.
Help the harness grow.
Everything here is Early Access. We are building the systems engineering harness in the open, together with the engineers who will use it. Bring a real evaluation case, show us where the MBSE flow breaks down and where AI assistance should help: backers directly shape what the harness becomes.
Start with the free 14-day Trial, become a backer with Pro to fund continued development, or involve us in a commercial Team or Enterprise deployment. What being a backer means.