How to Get Started with the Onshape FeatureScript MCP Server

What if you could get your CAD software to do exactly what you want it to do without becoming an expert in software programming?

The FeatureScript MCP Server from Onshape Labs gives you the tools to do so. The Model Context Protocol (MCP) server built by Onshape provides a direct connection from CAD to an LLM like Claude, ChatGPT, Gemini, or another MCP-compatible coding assistant.

Now, engineers can use their expertise to instruct an LLM to build, refine, test, and iterate on custom solutions using natural language instead of lines of code. From automating workflows or simplifying already-made scripts to building tailor-made design tools, the power to mold the Onshape experience is now in the hands of engineers and designers.

Here’s how to start experimenting with the Onshape FeatureScript MCP Server.

Setting Up the FeatureScript MCP Server

Onshape Labs is an early-access program giving Onshape users the ability to test and drive developing features and AI-powered tools. The FeatureScript MCP Server launched through Onshape Labs and is available through the Onshape App Store.

There are different ways to connect to various LLMs, but the following steps cover the basics.

1. Sign in to your Onshape account.
2. Subscribe to the FeatureScript MCP Server through the App Store.
3. Use your preferred MCP-compatible coding assistant.
4. Connect the FeatureScript MCP endpoint, which your coding assistant will relay to you.
5. Authenticate and verify the connection.

How Text Becomes CAD Automation

If you’ve never worked with an assistant like Claude or another LLM before, the interaction is simpler than it sounds – just write a request into a chat window. The assistant will take the request to generate FeatureScript, which can then be edited and executed from a Feature Studio.

The FeatureScript MCP Server is what makes this possible by creating a direct connection between the assistant and Onshape. Instead of copying code back and forth between a chat window and Feature Studio, the assistant can write FeatureScript directly into a document, review errors, iterate on the code, and inspect the geometry its code produces.

That means you’re no longer describing a model and hoping for the best. The assistant can work directly with the feature it’s creating and iterate based on actual results.

To get started, create a test document with a Part Studio and a Feature Studio. Then give the assistant a simple request and see what it generates. Add the feature to your Part Studio, test it, and evaluate the results.

If something doesn’t work as expected, just describe the issue the same way you would to a colleague:

• “The holes are on the wrong face.”
• “The pattern needs to continue around the entire part.”
• “This feature should work on multiple selections.”

The assistant uses that feedback to refine the feature and generate revisions. A few rounds of iteration are completely normal, and each revision helps shape the feature around your design intent.

Along the way, the generated code provides practical examples of how FeatureScript works in real-world applications, making it easier to understand and learn over time.

Start with a Problem Worth Solving

It’s tempting to try the most impressive thing possible, but a better approach is to focus on what would actually be useful.

Some initial ideas could start from thinking about repetitive steps or tedious parts of a design workflow. A few examples to start:

• “Build a bolt circle feature with inputs for face selection, hole count, pitch circle diameter (PCD), hole diameter, and angular offset.”
• “Create a mounting boss feature that places a boss on a selected face and automatically adds the hole and base fillet.”
• “Build a feature that generates a hole callout table from every hole tagged ‘critical’ in this part.”

Then test it out by adding the custom feature to the Part Studio and using it in a real workflow.

Or, try exploring customization that solidifies engineering knowledge. While the FeatureScript MCP Server dramatically lowers the barrier to creating custom features, it doesn't eliminate the need for engineering judgment. In practice, LLMs and AI-powered tools are best treated as collaborators rather than replacements. A custom feature can help encode standards, design intent, requirements, or company-specific documentation directly into the model.

For example:

• “Create a feature that flags any wall under our minimum thickness for injection molding.”
• “Build a feature that checks bolt-pattern edge distance against our fastener spec sheet.”
• “Build an Onshape custom feature for a welded plate lifting lug designed per ASME BTH-1 (Design of Below-the-Hook Lifting Devices).”

And always, always, validate and review for accuracy, functionality, manufacturability, and design intent. Engineering knowledge can’t be replaced – and Onshape Labs is early access, so treat generated output the way you'd treat a capable but new teammate’s first draft.

Develop FeatureScript Skills Along the Way

In many ways, the FeatureScript MCP Server is making learning FeatureScript more accessible.

Engineers with little to no coding experience can inspect generated code, ask questions about implementation decisions, and iterate through designs much faster than before. Even a basic understanding of FeatureScript will help troubleshoot issues, review generated output, and collaborate more effectively with AI.

The Onshape Learning Center and FeatureScript documentation remain valuable resources as your projects become more sophisticated.

Accelerate Development with Co-Complete

The FeatureScript MCP Server isn’t the only AI-assisted tool available through Onshape Labs. FeatureScript Co-Complete brings AI assistance directly into Feature Studio, helping generate code suggestions while you’re perfecting what’s already in the Feature Studio.

The two tools complement each other well. The MCP Server is ideal for generating new ideas, creating entire features, and experimenting with design automation. Co-Complete helps when the code is already generated.

Together, they make custom feature development more approachable than ever.

To start using FeatureScript Co-Complete, enable Onshape Labs in Account Preferences.

The Real Opportunity: Text-to-Code-to-CAD

The FeatureScript MCP Server makes it remarkably easy to start creating custom features. With the power of LLMs and AI in CAD, the real opportunity comes from identifying the engineering knowledge that’s currently trapped in spreadsheets, documents, repetitive workflows, and tribal knowledge – and turning it into reusable tools that anyone can use.

The setup only takes a few minutes. Figuring out what to build next is where things get interesting.