Know what any part should cost — and why.
Every part has a true cost. ARCNM computes it. Drop in a STEP file and a drawing; get back the process, the machines, the cycle times and the unit cost at every lot size — each figure traced to the physics that produced it. No black box. No guess wearing a confidence score. Physics is the ground truth; AI reads the drawing, recognises the geometry and calibrates on top of it. Built to be called: one REST call, an MCP server, your agents.
What ARCNM does
ARCNM is a physics-first should-cost engine for manufactured parts. Given 3D geometry (STEP/STP/STL), a 2D drawing (PDF), and a batch size, it returns an auditable should-cost — material, setup, cycle time, labour, overhead and unit price across a lot-size curve — and the full extracted part data behind it. First-principles physics is the ground truth; precision AI works in every layer, reading drawings, recognising features, planning the process and calibrating to your shop.
ARCNM by the numbers
- 54 operation types in the process taxonomy — 53 carry a DIN 8580 code; the bought-in catalogue part is the exception
- 26 machine classes for your fleet — 3-axis VMC to Swiss-type, laser to metal printer
- 536 material grades in the catalogue — 1,913 designations across 53 EN, DIN, ISO, AISI, ASTM, JIS and UNS systems — and the grades you register yourself beside them
- 62 workpiece material groups behind the cutting data — ISO 513 groups, each with its own cutting data
- 10 lot sizes priced per calculation — 1 to 2,500 pieces, with the optimum lot and the machine chosen at each quantity
- 38 + 37 heat treatments, coatings and surface finishes read from the drawing — ISO 4885 and ISO 2080-family vocabulary
- 22 manufacturability checks — per feature, graded by severity, before the shop floor
- 15+ ISO, DIN, VDI and ASME standards applied — from ISO 286 fits to ISO 4885 heat treatments
- 100 % of figures carry their provenance — which operation, which parameter, which source — on every line
- 1 REST call to price a part — upload-and-quote; seconds for a typical milled part
- 10 MCP tools for AI agents — explanation, drivers, breakdown, process plan, lot size, levers, compare, rank, environments, quota
- €0 for a failed run — pay only on a valid result; an identical re-run in the same month is never billed
Deep feature & PMI extraction
ARCNM recognises holes, pockets, faces, bosses, bends, slots, chamfers and threads from the 3D model, and AI reads dimensions, tolerances, GD&T and surface finish from the 2D drawing, checked against the ISO and ASME drawing standards (ISO 286, 1101, 2768, 5459, 13715, 21920; ASME Y14.5). Every feature and all PMI come back as structured, typed, versioned data — a first-class output you can reuse for master-data management, ERP/PLM backfill or similarity search.
Manufacturing-process physics
ARCNM identifies the process family from the geometry (milling, turning, sheet metal, tube) and plans the operations and machines — routing on to wire EDM, grinding, laser or waterjet where the features demand it. You get physics-grounded setup and cycle times per operation, then a lot-size curve with the unit price, the machine chosen and the break-points at every quantity — calibrated to your shop, with a confidence band that keeps every number honest.
Built for code and AI agents
One REST call (POST /api/v1/calculations/upload-and-quote) prices the part; an MCP server lets AI agents explain, compare and rank the result. Every calculation is fully auditable — from every euro back to the feature that caused it.
Automate the cost step in every workflow
The same engine is reachable from code and from agents: a REST API with an OpenAPI 3.1 spec to generate typed clients from, signed webhooks when a calculation finishes, OAuth for products and automation platforms, and an MCP server for AI agents. Price every incoming RFQ package on arrival, re-price each CAD revision in the change review, backfill ERP and PLM master data from the extracted part data, calibrate from your ERP actuals — with the documented recipes, not a connector project.
Plans that scale with your volume
Every plan includes a monthly calculation quota — one part costed in one environment; every re-pricing is a new calculation. Past your quota, you pay a flat per-calculation rate, and failed runs are always free. Start free with 10 calculations a month and no card; Pro is €199/mo (300 included), Scale €1,199/mo (2,500), and Enterprise is a contracted annual plan with contracted volume and a negotiated rate — talk to sales.
Frequently asked questions
How do I calculate the manufacturing cost of a part from a STEP file?
Upload the STEP model and, if you have it, the PDF drawing; choose a costing environment (your fleet and rates, or a ready-made reference) and a lot size. ARCNM recognises the features, plans the operations and machines, simulates setup and cycle time from physics, and returns an itemised unit cost across the lot-size curve — in the workspace, with one REST call, or from an AI agent over MCP.
Which manufacturing processes can ARCNM cost?
CNC milling and turning, sheet metal (laser, plasma and waterjet cutting, punching, shearing, press-brake, roll and tube bending), wire EDM, grinding, injection moulding and additive manufacturing from physics — plus heat treatment, coating, surface finishing and welding read from the drawing and priced as subcontract steps.
What is a should-cost?
A should-cost is the price a part ought to have when it is made efficiently at fair rates, computed bottom-up from material, cycle time, setup and overhead rather than taken from a supplier quote. ARCNM computes it from the CAD geometry and drawing, with every line traceable to the operation and parameter behind it.
Can an AI agent such as Claude or ChatGPT use ARCNM to price parts?
Yes. ARCNM ships an MCP server at api.arcnm.io/mcp-server/mcp. Any MCP client — Claude, ChatGPT, Cursor, Claude Code or a custom agent — can explain a calculation's cost, rank its drivers, optimise the lot size and compare parts or environments through ten typed tools; the part is uploaded and the calculation started with one REST call or in the workspace. Billing, credentials and tenant administration stay human-only.
Is ARCNM a black-box AI estimate?
No. First-principles manufacturing physics computes the cost; AI is used only to read drawings, recognise geometry, plan the process and calibrate the result on top of the physics. Every figure carries its provenance — which operation, which parameter, which source supplied it — and calibration to your shop adjusts the result, never the physics.
Can ARCNM integrate with my ERP, PLM or CAD system?
Yes — through the REST API, signed webhooks, OAuth and the MCP server; any system that speaks HTTP. The developer docs carry a field-mapping recipe for SAP S/4HANA, Autodesk Vault, Oracle Fusion and Odoo, and calibration accepts your ERP cost history directly. There is no connector to buy.
Does ARCNM train AI models on my parts and prices?
No. We never train on customer data — not identifiable, not anonymised, not aggregated, never across customers. It is contractually binding (Terms § 10b), enforced in the schema and guarded by tests. Your calibration and part data stay yours.