Product

Introducing moveOS: the operating system for agentic engineering

Bruno FincoGirish Radhakrishnan
Bruno Finco and Girish RadhakrishnanNovember 17, 2026 · 8 min read

Product · Engineering · AI Agents

One place for a company's data, tools, agents and people, inside its own cloud. Engineers build the Workflows and apps their team runs, from simulation and test to correlation and decision.

Developing a vehicle takes years and costs billions. Every manufacturer and supplier now has to speed up to stay competitive, often by 30% in the next year, and they have to do it with the engineers they have.

AI is already inside these companies, and program timing has not changed. Engineers use it for documents and scripts, and many teams run their own experiments, while the engineering itself is still done by hand. The reason is that designing a real vehicle takes far more than a general AI assistant can reach. It is the work of a whole organization, with specialized tools, test rigs, methods validated over years and many teams handing work to each other. The legacy tools behind that work were built for one engineer clicking buttons, and none of them was built for an agent.

For ten years I developed vehicles, and most of my weeks went into getting test data, simulation results and old reports into the same place and the same format, so the team could see the whole problem at once. Once that was done, the engineering decision, the part that needs an engineer, often took an afternoon.
Girish Radhakrishnan
Girish RadhakrishnanCEO and co-founder, MOVEdot

This is not a world AI agents can work in.

An engineering leader at a global automaker told us that what they need now is "one unified place to do agentic work, not a bunch of different POCs". So we built that place.

One place for data, tools, agents and people

moveOS is the one place for a company's data, tools, agents and people, inside the company's own cloud. It connects the design data, the simulation software, the test rigs and the proving ground, so AI agents can reach all of it and do the engineering work. Engineers build the Workflows and apps their team runs, and a colleague can open any result and see how it was produced.

One platform for testing, simulation and validation

Simulation

Agents run the simulations on the software the team has already validated, on the local workstation or the HPC cluster. They check every variant against the targets and come back with the best ones. In one study, agents ran over 5,000 multibody simulations in one night. At Pagani, the vehicle dynamics engineers set the targets and steered the study, and the agent compared about 6,000 damper variants in a few hours, summarized as the trade-off between comfort and track performance for each driving mode.

Our limits are no longer defined by throughput or time, but by imagination: our ability to conceive the right outcome, and discern the true value of the result.
Franco MorsinoHead of Vehicle Dynamics, Pagani
Pagani

Test

When a test finishes, its data is synced and indexed. The team's own processing and checks then run on every file as Workflows: fixed steps that convert the files, check the channels and build the standard plots and reports. Workflows run with no AI cost and call an agent when a result needs reasoning. HBK, which makes test and measurement equipment, turned an expert's troubleshooting method for electric-drive tests into Workflows and an app for the technician at the test stand. Up to 2 hours of expert troubleshooting became a 5-minute diagnosis.

It's taken years of my experience troubleshooting this data and put it in an app. Things I had a gut feeling for as an expert now have logic and science behind them.
Mitchell MarksDirector, Business Development Automotive, HBK
HBK

Correlation

Correlation is the comparison between a simulation model and a physical test, and it tells the team how far the model can be trusted. The agent compares the model with the test, finds what causes the difference, corrects the model and runs it again. This produced a full vehicle-model correlation in hours, a result also published by Canopy Simulations. We have partnered with VI-grade and HBK to help engineers compare simulation results with physical test measurements and use agents to investigate where the two differ. The same method applies to any simulation software and any test data a team uses.

Decision

Every result in the report comes with its evidence: the inputs, the method and the outputs. The engineer reviews it and approves it. Agents document every step, the decision and the reasoning behind it as the work is done, so nobody has to write it up afterward. The work can be inspected later, the context is carried from the target to the approval, and the next project starts from that record. A team can also compare its first results on moveOS with analyses it already trusts.

MOVEdot is so good, it gives you more time to ask questions and investigate things instead of doing the grunt work.
Matt SandersonSenior Engineer, Arrow McLaren
Arrow McLaren

A place to build

Every team works differently, so moveOS is a place to build. Engineers build their own Workflows and apps with agents, in hours for a first version, and what one engineer builds is something the whole team can run after. IT and digital teams build here too: the data pipelines and integrations the company needs, with usage, access and AI spending visible in one place. This is also why moveOS is used beyond vehicle programs, in tire development, in electric-motor test labs and in motorsport: any team that simulates, tests and correlates can build its own work here.

Pagani is an example. An analysis of the time lost in each braking zone was estimated at a week of one engineer's work. The engineers built it with an agent in less than a day, and it is now reused on every new session. The same team now compares thousands of design variants where it used to compare a handful.

Agents learn the company's methodology from its own reports, specifications and methods. The Knowledge Index makes those documents usable by agents, so a new investigation starts from what the team already knows. Agents then work as colleagues, each with a specific responsibility: for example, one keeps the test pipeline running, another maintains the apps, and a third runs model validation whenever new data arrives. Each company sets up the agents its own work needs. A vehicle dynamics engineer at Pagani describes it as "a new colleague who knows each type of analysis that exists in the world."

There is no AI magic in this. The expert defines the method and makes the decision, and agents run it at a scale no person has time for. What the company learns on one program is kept, and the next program starts from it.
Bruno Finco
Bruno FincoCTO and co-founder, MOVEdot

How moveOS is deployed

moveOS connects to what the company already has: the simulation software on its workstations and HPC cluster, the rigs and data loggers in the test lab, the folders and databases where test data is stored, and the systems that hold its reports and specifications.

moveOS works with the best AI model available, from any lab, including open-source models that run on the company's own infrastructure, so a customer is never tied to one model. Our benchmarks on engineering tasks show which model does each job best. Teams can also build their own benchmarks and evaluations on moveOS, follow how AI models improve and choose on quality and price. Other AI assistants can also reach moveOS through MCP (Model Context Protocol), the open standard that AI agents use to connect to tools.

moveOS runs in the customer's own cloud, set up with the IT team around its security and data requirements. IT controls who has access, sees how the platform is used and sets limits on AI spending. The enterprise deployment story describes this in more detail.

Our experts implement it with the team, on site. They connect the systems, build the first workflow with the customer's engineers and train the team on its own data and methods. They have done this engineering work themselves, in automotive and motorsport. At HBK, the troubleshooting system described above was built in two weeks with our engineers' support.

Where this goes

This is how engineering organizations will operate: one engineer with the reach of a team, agents that hold responsibilities next to the people, and each program starting from what the last one learned.

Bring us one workflow

The next generation of vehicles and machines will be developed this way, and the teams that start now will set the pace for everyone else. Which workflows slow your team down? Bring one to us, and we will set it up with your engineers, in person. Request a demo.