Swiss AI-native engineering company Neural Concept provides an intelligence layer for engineering, enabling companies to design and develop complex, high-performance products. Its AI-native platform embeds physics-aware intelligence directly into design and simulation workflows. The company this month launched MCP Bridge, the first capability to help AI agents operate commercial CAD, meshing and simulation tools in hybrid environments.
With this software toolset, AI agents carry out real engineering work inside the software used to design and simulate physical products, the way an engineer would.
Given that every physical product starts out life in engineering software, there may be something of a natural progression here.
What is an engineering “solver” model?
Large engineering organisations operate across complex software environments, licensing structures and security policies. In a typical simulation cycle, running the “solver” is the only step that needs no engineer. Note: a term that Neural Concept doesn’t stop to explain, we can define the solver as software that calculates how a physical design behaves under specified conditions, using mathematical models to simulate forces, heat, fluid flow or electromagnetism.
The work around it, from retrieving the right geometry and cleaning it up to meshing it, setting up the physics and feeding results back into the design, remains hands-on and often spans multiple tools. With MCP Bridge, Neural Concept gives agents both access to these tools and the engineering know-how to use them, while companies retain control and ownership of their programs and data.
“For the first time, the software that shapes every physical product can be operated by AI at enterprise scale,” said Pierre Baqué, CEO and co-founder of Neural Concept. “The real opportunity is to give engineers more leverage, so they can take on harder problems and solve them faster.”
This capability comes at a time when manufacturers face growing pressure to shorten product development cycles, while engineering experts remain a scarce resource. When agents handle approved tasks inside design and simulation software, every hour they take on is an hour handed back to performance optimisation. In one customer case, an engineer using MCP Bridge cleaned up the geometry of a wheel rim in approximately three minutes, compared with a day using the manual process.
Baqué says that the engineering workflow for a physical product moves from design through model preparation to simulation. Built on the open Model Context Protocol, MCP Bridge lets agents carry work from CAD to mesh (A mesh divides a 3D engineering model into small elements so simulation software can calculate how forces, heat or fluids affect the design.) to solver across a company’s Windows workstations and Linux clusters, without engineers leaving Neural Concept.
Neural Concept users at TRATON and OPmobility are already putting AI agents to work in engineering. Their engineers point to the same next step: giving those agents the ability to operate the CAD and simulation tools their teams already rely on.
What next-generation engineering looks like
“AI agents have completely changed the way I work,” said Ciro Alosa, Electromagnetic Design Engineer at TRATON and Neural Concept user. “I can run more work in parallel, and going from a research paper to a validated script in our simulation toolchain now takes a couple of days. The next-generation engineer focuses on planning and evaluation, managing teams of AI agents working towards a goal. When agents can operate the engineering tools we already use, nothing is out of reach anymore.”
“At OPmobility, we are exploring how AI agents can operate the CAD and simulation tools our engineers already use,” said Yoann Eulalie, R&I Group Leader at OPmobility. “The aim is to give engineers more time to explore design options, while they keep full control of the decisions that require engineering judgment.”
Because each stage of an engineering workflow requires different capabilities, the top-performing model varies by task. Neural Concept benchmarks models on real engineering tasks, including CAD editing and meshing, so companies can select the model most suited to each task as new models become available.
MCP Bridge is available now to Neural Concept’s customers and extends each company’s Engineering Intelligence Layer.
What this means for developers
Neural Concept’s MCP Bridge extends agentic AI beyond software development and into the engineering applications used to design physical products.
For developers, the significance here lies in connecting AI agents to established commercial tools through the open Model Context Protocol (MCP). Rather than simply generating scripts or suggesting design changes, agents can execute tasks across CAD, meshing and simulation workflows, spanning Windows workstations and Linux clusters.
Developers and engineering teams can automate repetitive preparation work, integrate existing toolchains and select models suited to specific engineering tasks. The challenge that software engineering teams will have is ensuring agents operate within enterprise security policies, while developers themselves will also still need to retain control over validation and design decisions.