Turning AI in CAD into Engineering ROI: Make Every API Call Count

Working with AI in Onshape is a great way to build custom engineering tools faster, and it’s worth knowing how API usage works along the way. AI clients that connect to Onshape through an MCP server may make more calls than a traditional script in a self-improving workflow, so your allocation may go quickly as you get started. With a clear goal, a focused scope, and a budget, you can put that allocation to work where it delivers the most.

Turn Engineering Knowledge into Reusable CAD Features

For decades, CAD vendors have built general-purpose systems, while much of a company’s most valuable engineering knowledge has lived outside them in standards, spreadsheets, and the heads of experienced engineers. Capturing that knowledge inside CAD has traditionally required specialized programming expertise.

FeatureScript MCP Server from Onshape Labs changes that.

Connect an AI client such as Claude Code, Codex, or GitHub Copilot, and an engineer can describe a needed capability in plain language. The AI writes, tests, and debugs the FeatureScript, resulting in a native, parametric feature in Onshape.

That’s the opportunity. Generating geometry saves work once. Capturing engineering logic saves work every time it’s reused. Once validated, a custom feature runs inside Onshape like any other, with no additional API calls or AI tokens, and it can be shared across the whole company. AI helps build the tool once, then engineers can use it repeatedly without using more AI tokens or Onshape API calls. Over time, those features become a library of company-specific capabilities that reflect how your teams design.

Early users are already reporting significant time savings. One early adopter built a custom feature to automate a repetitive modeling task that takes about 15 working hours today. They expect to bring that down to roughly three hours, even with time set aside for troubleshooting.

In another case, a user worked for 7 hours leveraging the FeatureScript MCP Server, about 2,500 API calls, on what would otherwise have taken about a month. The team calls it a huge, repeatable win, placing Onshape as the backbone of their AI adoption strategy.

With FeatureScript MCP, engineers can create:

• Standards-based features that every designer applies consistently
• Product families that rebuild instantly when a parameter changes
• Built-in checks that keep designs aligned with company rules
• Calculators that turn live geometry into engineering answers

FeatureScript MCP Server is available through Onshape Labs. As with any new tool, review each custom feature before adding it to a production workflow.

Get More Value from Every API Call

When an AI client is connected through MCP, it decides which requests to make and how often. It gathers context, uses tools, checks its results, and tries again. The following practices keep that effort focused on your goal.

1. Plan Before Building

For larger tasks, ask the AI to propose an approach before it changes anything: “Review Part Studio 3 and propose a FeatureScript approach for identifying undersized fillets. Don’t modify the model until you’ve explained the plan and validation steps.”

2. Set a Clear Objective, Scope, and Context

Replace “improve this design” with “design the garden hose thread in accordance with ASME V1.20.7 standards while preserving design intent and parametric flexibility.” Point the AI to specific parts rather than the whole document, and share requirements, constraints, design intent, and known issues up front so it doesn’t spend calls discovering them.

3. Break Big Projects into Stages

Import and organize CAD data, create mate connectors, generate the FeatureScript, then validate the geometry. Each stage gives the AI a clearer direction.

4. Give the AI an API Call Budget

A budget keeps the AI’s thoroughness pointed at the goal. Add it to your prompt: “Complete this task in 100 API calls or fewer. Check in with me at 75, and if you reach 100, stop and summarize your progress.” Treat the AI’s count as a helpful guide, and confirm actual usage in your Developer settings.

5. Monitor Usage and Measure the Outcome

Watch API usage directly in Onshape as you develop a new workflow, especially before repeating it or expanding it across a team. Then weigh it against what the automation delivers: less repetitive work, standardized modeling tasks, and engineering knowledge that carries across projects. API usage can be found under Account, then Developer, or simply ask for updates through the AI client.

What to Know About API Allocations

Every Onshape subscription includes unlimited time in the design workspace. You can open as many documents, regenerate as many features, run and re-run simulations, and collaborate as much as you like. API allocations apply to automated work, where a script, integration, or AI can make thousands of requests in the time it takes a person to perform a few clicks. Counting that use separately allows seat pricing to stay flat for everyone.

Onshape’s own MCP servers are counted like any other API client, since the AI decides which requests to make. API calls are also separate from AI tokens, which measure how a language model processes prompts and responses. Read “AI Tokens in CAD: What Engineers Using LLMs and Coding Agents Need to Know” to learn about the difference.

Annual allocations vary by plan. Head to the Onshape API documentation for a subscription breakdown and more information on additional capacity.

Build Once, Benefit Often as AI and Onshape Evolve

AI workflows are evolving quickly, and Onshape’s cloud-native architecture is built to keep pace. In the meantime, focused prompts, thoughtful planning, and reusable features help every API call go toward something your whole team can use.