Authority
A named person approves the actions that matter and can stop them.
Company overview / first introduction
Neura Parse is a UK software company with three products: NowFlow for AI agent workflows, NeuraOS for robots and edge devices, QFlow Studio for quantum experiments.
The problem we work on
When AI touches a real operation, the team needs to know who approved the action, what limits applied, what changed, and whether they can retrace it later.
That is a systems question, and it is the one we build for.
How we think about it
The same three ideas run through every product we make.
A named person approves the actions that matter and can stop them.
It is always clear which agents, models, devices, and safety limits are in play.
Every decision, run, exception, and result is kept so a reviewer can check it later.
Our three products
They share the same principles, but each is a separate product, sold and run on its own.
NOWFLOW
Build AI agent workflows on a visual graph, with approval steps and stop points that people control.
NEURAOS
An operating-system foundation for robots and edge devices that run AI, with safety limits kept separate.
QFLOW STUDIO
A web app to design quantum circuits visually, generate code, run them, and keep a record others can review.
Commercial product / available today
NowFlow lets a team build AI agent workflows on a visual graph. Each agent has a set scope, people approve the key steps, and every run is recorded.
A named person keeps the final say. Robot and device safety systems keep their own authority.
Commercial product / public reference build
NeuraOS runs AI close to the hardware, for perception, planning, and inference, while a separate deterministic safety layer can pause, override, or stop the machine.
Public reference build rc.8, tested in simulation only. Physical operation is not yet cleared. The public repository holds only a README.
Commercial product / live web app
QFlow Studio is a web app for quantum work. Design a circuit visually, get Qiskit, Cirq, or OpenQASM code, run it on a simulator, and keep the history and exports.
Routes to real quantum hardware are labelled by status (beta, preflight, planned). A provider logo is not an endorsement.
What they have in common
Each product covers a moment where an instruction turns into real work: a workflow, a machine, or an experiment.
Who may approve what, what is allowed, and what happens on failure.
A workflow may approve a task. The robot's safety layer can still refuse.
Code, route, run, result, and known limits are kept together.
Applied research programme
It studies how robots, vehicles, and drones from different makers can share one picture of the world, take orders that can be checked, and keep a person in charge.
The core must work with ordinary computers and without a constant cloud link.
Quantum methods are tried only on slow, non-safety tasks, against a classical baseline.
What is public is the architecture and proposed pilots, not customer deployments or certifications.
Quantum research, all public
All three are public. One is about decisions, one is infrastructure, one is a learning library.
QANTIS
Can a quantum computer help a robot decide under uncertainty? Two public papers on arXiv so far.
qmesh
A shared representation for quantum programs with signed records. Apache-2.0, 157 passing tests.
QMANN
Quantum-classical neural networks in three modes, with a cost estimate before any paid hardware run.
What is public today
These are public engineering facts. They say nothing about customers, revenue, deployments, or certifications.
arXiv:2603.00785 and arXiv:2607.06760
Open-source repository, phases 1-6
Hugging Face inventory as of 4 Jul 2026
NowFlow, NeuraOS, QFlow Studio
Where this is useful
Infrastructure inspection, remote field work, robotics teams, research labs, and quantum groups all need the same thing: clear limits and a record they can check.
Illustrative scene. Not a customer deployment.
How we work with you
A first engagement is deliberately small. It ends with a clear decision to stop, adjust, or expand.
Pick one decision that matters, who owns it, and what a safe stop looks like.
Map the workflow, the systems involved, who approves, and how to recover.
Run the normal case and the failure cases in the agreed setting.
Review the record together and decide whether to widen the scope.
The company in one table
We keep the categories separate on purpose: what is a product, what is research, and what is still proposed.
Next step
In one session we map who approves it, which systems touch it, and what record you need. Then we decide what to build.
nowflow.io / qflow.studio