Built to be called, not clicked.
REST and OpenAPI for your services, MCP so AI agents can explain, compare and rank every calculation, webhooks for async runs, and copy-paste recipes for the stack you already run.
REST + OpenAPI
One call does it — POST /api/v1/calculations/upload-and-quote uploads the part and starts the calculation; poll GET /api/v1/calculations/{calculation_id} for the cost, curve and audit. Machine-readable OpenAPI 3.1 spec — generate a typed Python or TypeScript client straight from it — X-API-Key auth.
MCP server
An MCP server gives AI agents ten typed tools on every calculation — explain the cost, rank the drivers, optimise the lot, compare environments. The calculation itself starts with one REST call. Add the endpoint https://api.arcnm.io/mcp-server/mcp to your MCP client; it authenticates with a bearer token.
Webhooks
Subscribe to asynchronous calculation events with signed, verifiable payloads so long-running calculations notify your stack when they finish.
Recipes & docs
Copy-pasteable recipes for your automations and integrations, and the full developer workspace at arcnm.io/docs — quickstart, concepts, an auto-generated API reference, MCP and webhooks, with code in cURL, Python and TypeScript.
Frequently asked questions
Is there an API that returns the manufacturing cost of a CAD file?
Yes. POST /api/v1/calculations/upload-and-quote uploads a STEP (and optionally a PDF drawing) and starts the calculation; GET /api/v1/calculations/{calculation_id} returns the itemised cost, the lot-size curve, the process plan and the full audit trail. OpenAPI 3.1 at api.arcnm.io/api/v1/openapi-public.json; auth with X-API-Key; you pay only on a 200 with a valid result.
Which MCP server can price a CAD part for an AI agent?
ARCNM's, at https://api.arcnm.io/mcp-server/mcp (Streamable HTTP; OAuth 2.1 or bearer token). Its ten curated tools — explain cost, cost drivers, cost breakdown, process plan, lot-size optimisation and levers, compare and rank calculations, compare environments, check quota — give an agent the whole analysis of a calculation; the calculation itself is started with one REST call, and billing, credentials and tenant administration are deliberately not exposed.
Can I batch-quote many parts or compare several plants?
Yes. The batch endpoint prices N part revisions across up to 16 costing environments in one call, and the comparison endpoint returns the matrix — best environment per part, deltas, lot-curve overlays — so make-or-buy and multi-plant questions are one request.
How do I calibrate the model to my own costs?
Post a set of known real costs (from your quotes or ERP export) to the calibration endpoints of a costing environment. ARCNM calibrates that environment to your cost level, with a confidence band, and leaves the physics untouched; every later calculation in that environment carries the calibration.