5 formas en que los equipos pueden aprovechar el servidor Onshape FeatureScript MCP

AI in CAD won’t remove engineering decisions, the creativity, nor the expertise required to build great products.

What it is doing is removing some of the most frustrating and tedious parts of product development, while creating CAD workflows tailor-made to teams, feature customizations that speed up iteration, and uncovering optimizations quicker than ever before.

Enabling this breakthrough is the FeatureScript MCP Server by Onshape Labs. The server connects Onshape, PTC’s cloud-native CAD and PDM platform, to large language models (LLMs) such as Claude, Microsoft Copilot, Gemini, and Perplexity, enabling engineers and designers to describe what they want in plain language and have AI help create, test, debug, and refine custom CAD functionality.

The FeatureScript MCP Server enables engineers who aren’t as familiar with coding custom features to create the tools they need. It also speeds up tool development for coding experts.

Already being put into practice, take a look at five ways the FeatureScript MCP Server can enable better product development through a text-to-code-to-CAD workflow.

Use Cases: FeatureScript MCP Server

The scope is vast and the possibilities numerous. To start, here are five use cases:

Build Geometry With Standards Encoded

Engineering teams can spend hours recreating the same specialized component over and over again, be it a standards-based thread, catalog-driven part, or company-specific design features.

With the FeatureScript MCP Server, teams can describe the intent and associated requirements in natural language to generate a fully parametric custom feature that becomes part of a team’s Onshape toolset.

Take, for instance, an engineer describes a garden hose thread that’s built to ASME V1.20.7 standards. The LLM will build out the FeatureScript code for an Onshape custom feature. Or perhaps the engineer wants a roller chain sprocket generator pulling from online catalogs with configurable strand counts and automatic manufacturing annotations.

These are real features – not macros or one-off scripts – that can be configured, edited, patterned, and shared. They’re able to carry requirements from an LLM request into Onshape to show model-based definition (MBD) annotations.

The tedium of rebuilding component specifications vanishes, and iteration happens through simple parameter adjustments rather than feature recreation.

Create Reusable Feature Components

Those familiar with using FeatureScript know that custom feature development reveals recurring patterns. Color maps, material lookups, geometric calculations, and utility functions often get rebuilt from scratch, creating extra work and inconsistencies between projects.

So, instead of rebuilding these utilities each time, ask the LLM to generate reusable function libraries that solve these cross-cutting concerns. The MCP server intelligently inserts these libraries into the Feature Studio so subsequent features reference them by name.

For example, a color spectrum library with switchable gradients, utilities for material property retrieval and dimensional lookups, or spatial query functions can all accumulate into a personal toolkit that grows more valuable with each project.

Develop Sophisticated Features for Specialized Workflows

Bridge the gap between what’s needed in an engineering workflow and what a CAD system offers out-of-the box.

For instance, building a blend corner feature that takes intersecting fillets and creates smooth transitions between them, rather than sharp corners. Or developing a niche feature specific to the company standards and designs. These capabilities encode domain expertise directly into the CAD system rather than living as workarounds or external processes.

LLMs excel at algorithm development, so they can optimize for performance, quality, robustness, or readability based on requirements, so specialized workflows stop being painful and become embedded in the design process.

Wrangle Code through Featurization

Complex geometry and custom feature code accumulate over the years. Simplifying and refactoring into one custom feature is now possible through LLM-driven FeatureScripting.

In this case, a piston model refined through meticulous reverse-engineering contains 60 features, parametric expressions, and decades of engineering intent encoded across thousands of lines of code.

That existing work doesn’t need to be abandoned but can be compressed and optimized. Just feed the LLM the code and ask it to consolidate those 60 features into a single parametric feature.

Then it can be used and shared with all native feature operations working seamlessly. Or extract just the useful pieces by grabbing a few features from a larger model, then ask the LLM to turn them into a standalone reusable feature. Now, monolithic models can become modular and shareable.

Advanced Features with Integrated Analysis

Some design decisions require analysis, simulation, standards checks, and calculations before you can move forward. Think thermal analysis, structural simulation, standards compliance checking, system-level optimization.

Instead of jumping between CAD, spreadsheets, reference materials, and specialized analysis tools, teams can build custom applications that bring those workflows directly into Onshape.

A custom application, such as an FEA beam analysis tool, can let users select curve geometry, set material properties and load cases, then converge on deformed shapes and frequency response all within a Part Studio.

Or, create a heat exchanger design system that generates complete assemblies based on ASME standards, configures shell-and-tube layouts, and automatically calculates thermal performance with detailed output tables.

These applications represent a new category of CAD capability: Geometry creation that’s intelligent, purposeful, and outcome-driven.

Once built, they’re deterministic and efficient because the LLM did the heavy lifting during development and subsequent runs consume no AI tokens.

Engineering analysis that would normally demand switching between specialized software and manual data transfer now exists directly in the CAD workflow, surfacing optimization opportunities faster than ever before.

What Will You Build?

These are just a few examples to get the idea juices flowing.

The FeatureScript MCP Server closes the loop between AI-generated code and working CAD functionality. It allows engineers to focus on defining requirements and engineering intent while AI helps with the implementation.

That means fewer barriers between an idea and a working solution, which just might be one of the most exciting applications of AI in CAD yet.