Romanian service provider 3Descu went live on October 18, 2026 with an MCP server that lets ChatGPT, Claude, Cursor and VS Code pull manufacturing quotes directly from its existing pricing engine, no account or API key required.
An engineer drops a publicly reachable HTTPS link to an STL or OBJ file, names a material and a quantity, and the assistant returns a price. The server is listed in the official MCP Registry as io.github.cadi27tech/3descu-quote and exposes four functions: manufacturing capabilities, available materials, model analysis, and price estimation. The model-analysis function reads part volume, bounding dimensions, surface area and mesh watertightness, feeding those into the same calculation engine that runs the company’s website, so AI-generated quotes match the ones a customer would get by hand.
| Detail | Figure |
|---|---|
| Live date | October 18, 2026 |
| MCP functions exposed | 4 |
| FDM/SLA printers in-house | 200+ |
| Orders completed | 1,200+ for 450+ clients, 10+ EU countries |
| Germany’s share of orders | ~45% |
| Founded | June 2022, with 5 printers |
The integration stops short of placing orders. Once priced, the assistant hands back a link to 3Descu’s quotation page, and the company’s production team still reviews every file before manufacturing, with prices subject to change if mesh geometry or other issues surface during that check. Founder Adrian Căpraru framed the pitch as getting “a real price from a real factory in the same chat, without filling in a single form.”
The move sits apart from how rivals have used AI in quoting. AMFG’s Sentinel AI, introduced in July 2025, reads 2D drawings for tolerances and surface finish in under three seconds, and DigiFabster’s Quote Agent, launched December 2024, targets RFQ processing. 3Descu isn’t interpreting drawings or managing requests; it’s making an already-built pricing engine reachable from inside a chat window, which is a smaller claim but a more immediately testable one.