AI DFM that measures the part, not guesses.
Ask a language model whether a part can be made and it will answer, confidently, from a description. It can't measure a wall, reach into a pocket or count setups. Protyus does the measuring: it reads the exact geometry in your STEP file and reports what a manufacturer would push back on, where it is on the part and the change that fixes it. Then AI does what it's good at: answering your questions and acting on the result.
Design risks, found on the part itself.
- Findings you can triage
- Each one is a Blocker, a Risk or something to consider, shown on the part with why it matters and the change that fixes it.
- A score for every part
- A score out of 100, and for a machined or sheet metal part a verdict: whether it can be made as modeled.
- The same answer every time
- Run a part twice with the same settings and you get exactly the same answer, to the last number, whoever runs it.


An opinion, or a measurement.
Tools sold as AI DFM work in one of two ways. In the first, a model looks at a picture, a drawing or a description of the part and writes a review. It's quick, but it can't measure anything, and asked the same question twice it can give two answers. In the second, software measures the model itself and holds each measurement to a limit.
Protyus is the second kind, with AI on top. Its findings come from the part's geometry, checked against established manufacturing practice and your own limits, and each one names the measurement and the limit it was held to. AI never decides whether a feature can be made. It explains what was found, answers what you ask and carries the result into a quote.
Each process, by its own rules.
Protyus covers CNC machining, sheet metal fabrication and injection molding, with more process groups on the way.
CNC machining DFM
Sharp inside corners, thin walls, deep holes and features a tool can't reach, with the setups on 3-axis, 4-axis, 3+2 and 5-axis machines.
Sheet metal DFM
Bend radii, flange lengths, holes near bends and reliefs, the flat pattern as a DXF, and a bend order a press brake can make.
Injection molding DFM
Wall thickness, draft, ribs, bosses and undercuts judged against your resin, with the parting line and the slides the mold needs.
Ask about the part, or let your agent work on it.
- Protyus AI
- The assistant in the app answers questions about your part in plain words, from its analysis: what 50 will cost, and what a shop will push back on.
- Your own agent
- Protyus is an MCP server. Claude, ChatGPT, Cursor and other MCP clients upload a part, run the analysis and read back the setups, findings, cycle time and cost.
- Your own systems
- Every MCP tool is also a REST API operation, described in a public OpenAPI document, with results as structured data with units.
- Your files stay yours
- We don't train AI models on your parts, and Protyus AI never sends your STEP file or its 3D shape to an AI model.
Customer RFQs · Haas UMC-500 · 6061-T6 · qty 50
It can be made as modeled, in two setups on the Haas UMC-500.
A shop would raise two things:
- The slot (1). It's 3 mm wide and 60 mm deep, too deep for an end mill, so it needs a slitting saw.
- The R3 corners. They match a Ø6 cutter exactly. Go slightly larger so the tool cuts them on an arc.

$69.94 each, $3,496.94 for 50, in 6061-T6 on the Haas UMC-500, at 11.8 min a part.
More on what Protyus does.
DFM analysis
What a DFM check finds, how findings are ranked, and how it leads to a cycle time and a cost.
MCP server and API
Connect Claude, ChatGPT or your own agent, and what it can do once it's connected.
What is AI DFM?
AI DFM is a design for manufacturability review done by software instead of in a meeting: you give it a CAD model and it reports what will be hard or impossible to make, and what to change. Some tools have a language model review a picture, a drawing or a description of the part. Protyus measures the geometry in the STEP file, and uses AI to explain what it found and act on it.
Can ChatGPT or Claude do a DFM review?
On their own they can only guess: a language model can't measure a part, so it can't tell whether a tool reaches a feature or how many setups the part takes. Connected to Protyus through its MCP server, they read the real findings, setups, cycle time and cost for your part and build on them. The pricing page says which plans include the MCP server.
Is the DFM analysis itself done by an AI model?
No. The findings come from measuring your model and checking it against established manufacturing practice and your own limits, so the same file gives the same answer every time, and each finding names the measurement and the limit it was held to. AI is how you ask about the result, in the app or from your own agent.
Which files does it read?
STEP files (.step or .stp), which every CAD system can save.
Which manufacturing processes does it check?
Protyus covers CNC machining, sheet metal fabrication and injection molding, with more process groups on the way.
Does Protyus train AI on my parts?
No. We don't train AI models on your parts, your results or your settings, and Protyus AI never sends your STEP file or its 3D shape to an AI model.
Give your AI a part it can measure.
Protyus is opening to shops, suppliers and design teams by invitation. Ask for early access, or write to us with any question.