Direct answer: A general LLM does not safely control SolidWorks by conversation alone. It needs a tool layer—such as a reviewed add-in, macro service, API adapter or controlled desktop worker—that converts bounded instructions into supported operations. The adapter must validate arguments, restrict file and data access, capture SolidWorks return states and refuse unsupported actions. The LLM can help interpret and plan; the CAD runtime remains the source of execution facts.
How the controlled workflow works
- Define a small command contract with typed parameters, allowed files, version requirements and explicit denial rules.
- Validate the model’s proposed action against engineering state and permissions before any CAD mutation occurs.
- Execute through documented SolidWorks APIs or a controlled UI worker and record the actual return values.
- Read the native document after execution; never let the language model declare success from its own plan text.
What evidence should a reviewer inspect?
| Review item | What it can establish |
|---|---|
| Typed tool contract | Prevents free-form text from becoming unrestricted desktop or file-system control. |
| Runtime event log | Distinguishes proposed actions from calls accepted and completed by SolidWorks. |
| Post-action document facts | Verifies components, features, mates and errors from the CAD application. |
Where this answer stops
Generated VBA, C# or API code may be a useful prototype, but it is not production-safe merely because it compiles. Hallucinated method names, unsafe paths and partial document state need independent controls.
Human engineering authority
An authorized engineer or administrator approves tools, scopes and changes; production release remains outside the LLM.
Related ways people ask this question
- Can ChatGPT control SolidWorks?
- How do I connect an LLM to SolidWorks?
- Can Claude or another AI operate SolidWorks?
References and claim boundary
- SOLIDWORKS API Help 2026 — IAssemblyDoc
- SOLIDWORKS Help 2026 — SOLIDWORKS Labs (Beta)
- NIST AI 100-1 — Artificial Intelligence Risk Management Framework
Limit: External sources support the named standards, platform features or governance concepts. They do not endorse MST or independently validate MST product performance. MST implementation statements remain bounded by the current public evidence status.
MST Engineering AI. “Can ChatGPT or an LLM Control SolidWorks?.” Engineering Questions Q04. Reviewed 2026-08-15. https://mst-us.ai/questions/can-chatgpt-llm-control-solidworks/
NEXT STEP · NON-CONFIDENTIAL FIRST
Review an agent integration boundary
Describe the workflow and decision boundary without uploading protected P&IDs, CAD, BOM, PDM data, credentials or customer identities. Confidential exchange begins only after identity, NDA and scope review.