Neural Concept launches MCP Bridge to let AI agents operate CAD, simulation tools

The Switzerland-based AI-native engineering company said the bridge is intended to help large engineering organizations automate complex software environments used to design and simulate physical products, a central part of manufacturing and proving prototypes before production.

In a typical simulation cycle, running solvers is the only step that does not require human engineers and is already largely automated, the report said. The rest involves retrieving the correct physical geometry, cleaning it up and turning it into a virtual mesh.

Founded in 2019, Neural Concept serves original equipment manufacturers and engineering organizations in automotive, aerospace, energy, consumer, electronics, semiconductors and defense. It integrates an AI-augmented engineering platform and deep learning into CAD and computer-aided engineering workflows.

With MCP Bridge, the platform gives AI agents access to the tools needed for those tasks and the engineering know-how to use them, while companies retain control and ownership of their software and data, according to the company.

“For the first time, the software that shapes every physical product can be operated by AI at enterprise scale,” Chief Executive and co-founder Pierre Baqué said in the announcement.

Baqué said the goal is to reduce tedious work so engineers can take on harder problems and work faster. The company pointed to manufacturers seeking to shorten development lifecycles with fewer engineering experts available, according to the report.

Neural Concept said that if an agent can handle an approved task in design and simulation software, each hour it takes is an hour returned to an expert for more critical work.

In one customer case, an engineer using MCP Bridge cleaned up the geometry of a wheel rim in three minutes, compared with a day using a manual process, according to the company. That preparation is needed to move a physical product into a virtual model and then into simulation.

The process of virtual model preparation and simulation shifts engineering from reactive physical trial and error into proactive, data-driven digital environments by creating digital twins. Those twins can be simulated in high fidelity only if they are cleaned up to act, look and react like their real-world counterparts under real-world physics before manufacturing, the report said. The preparation draws on spec sheets, materials guidelines and large amounts of data and can be tedious, long work requiring extensive validation.

Once a model is ready, engineers can test thousands of design scenarios at the same time in virtual labs instead of waiting for physical prototypes. But getting one set up and into a rough draft that an expert can refine can take weeks or more; an AI agent working in computer-aided engineering software can cut that time to hours, leaving the engineer to fine-tune the result, according to the report.

“AI agents have completely changed the way I work,” said Ciro Alosa, an electromagnetic design engineer at Traton SE, a Volkswagen Group subsidiary that manages heavy truck, bus and light commercial vehicle brands. “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.”