Put cost into every workflow — ERP, PLM, CAD and AI agents included.
Built to be called, not clicked. One REST call prices a part. A signed webhook reports it done. The MCP server lets an AI agent upload a part, start the calculation and read, explain and compare every result. OAuth lets a product or automation platform act for its users. The cost step lands where your team already works. That can be the ERP quote, the PLM change review, the CAD session or the procurement agent. You connect it with documented recipes, not a connector project.
What can I automate with ARCNM?
- Price every incoming RFQ package on arrival. The batch endpoint prices up to 500 part × environment cells in one request, up to 16 costing environments per part. The comparison endpoint returns the best environment per part, with the deltas. So triage happens before anyone opens a drawing.
- Design-to-cost in the change review: re-price each CAD revision from the PLM or CAD system. Post the cost delta back to the ticket. So a tolerance or material change is priced while it is still open.
- Fill ERP and PLM master data: every calculation returns the extracted features, dimensions, tolerances, treatments and materials as typed data. Write them to the item master instead of keying them.
- Calibration from your ERP actuals: post known real costs from an ERP export to a costing environment. Your actuals calibrate it to your plant, supplier or region. A preflight shows what the export covers before anything is billed.
- Make-or-buy and multi-plant sourcing: run the same part across your plants' and suppliers' environments. The comparison endpoint returns the winner per part, the deltas and the totals per environment.
- Agent workflows: an AI agent in Claude, ChatGPT, Cursor or your own stack starts a calculation itself over MCP. It gets an upload link for a person to drop the STEP or PDF file on. Or it uploads the file directly. Then it starts the pricing. It explains the cost, ranks the drivers, optimises the lot size and compares options through the MCP tools. Billing, credentials and tenant administration stay with people.
Eight numbers you can plan against
- 1 REST call to price a part — POST upload-and-quote, OpenAPI 3.1, one spec for any client generator
- 15 MCP tools for AI agents — the default list. Upload and price, upload link, re-price, status, list, cancel, explanation, breakdown, process plan, what-if, lot size, levers, compare, environments, quota.
- 9 webhook events, signed — calculation, extraction-budget, wallet and billing signals
- 536 material grades, addressable by any designation — EN, DIN, AISI, ASTM or JIS code, or your own house name
- 100 material references resolved per call — any EN, DIN, AISI or house spelling to one canonical grade. 1.0038, S235JR, St 37-2 and Fe 360 B are one.
- 500 calculations per batch request — part × environment cells, up to 16 environments per part, compared server-side
- 512 MB per upload — STEP incl. AP242 PMI, STL, PDF and raster drawings
- €0 for a failed run — pay on success. You can set spend caps per organisation and per API key.
How does ARCNM connect to ERP, PLM and CAD systems?
Through standard interfaces your team already knows. The REST API comes with an OpenAPI 3.1 specification. You generate a typed client from it in any language. Signed webhooks report the events calculation.succeeded and calculation.failed. So an integration is event-driven, not polling. OAuth 2.1 with PKCE lets a product or automation platform act for many users ("Connect to ARCNM"). The MCP server connects AI agents and IDEs. The developer docs carry a field-mapping recipe for SAP S/4HANA, Autodesk Vault, Oracle Fusion and Odoo. The recipe uses idempotent upserts. The same pattern fits any PLM or quoting system that speaks HTTP. We do not sell connectors. The recipes are a few dozen lines, and they are yours.
Which integration path should I use?
| Path | Best for | Authentication | How the result arrives |
|---|---|---|---|
| REST API + OpenAPI | Scripts, ERP / PLM middleware, CI jobs, quoting tools | API key with scopes | Poll the calculation, or subscribe to the webhook |
| Webhooks | Event-driven sync into ERP, PLM or a data warehouse | Signed payloads, verifiable per delivery | Pushed on calculation.succeeded / calculation.failed |
| MCP server | AI agents, copilots, IDEs, custom agent stacks | OAuth 2.1 or bearer token | Typed tool results inside the agent's conversation |
| OAuth "Connect to ARCNM" | Products and automation platforms acting for many users | OAuth 2.1 + PKCE, tenant-scoped tokens | Same REST API, on the user's behalf |
What does automation cost, and can it run away?
You pay only for a calculation that returns a valid result. Failed runs are free. A re-run within a billing month returns the first result and is not billed. This applies to the identical part in the same environment with the same configuration. Monthly spend caps per organisation and per API key pause further runs when reached. So a batch job, a webhook loop or an agent can never bill more than you allowed. Rate limits are documented per plan. Every paid plan includes unlimited API keys and seats.
Frequently asked questions
Can ARCNM connect to SAP, Oracle, Odoo, Autodesk Vault or my PLM?
Yes, through the REST API and webhooks — any system that speaks HTTP. The developer docs include a field-mapping recipe for SAP S/4HANA, Autodesk Vault, Oracle Fusion and Odoo. The recipe uses idempotent upserts. A PLM or quoting tool follows the same pattern. There is no vendor connector to buy and none is needed.
Do you have apps for Make, Zapier or n8n?
For n8n, yes: ARCNM is a verified node in n8n's integration directory (https://n8n.io/integrations/arcnm/). It uploads a CAD file and starts the quote. It also reads calculations, the cost sheet, parts and the remaining quota. Make and Zapier have no listed app yet. They connect through the OAuth 2.1 "Connect to ARCNM" flow, and their generic HTTP modules call the REST API. The quickstart request is a single multipart call.
How does an AI agent run a quoting or sourcing workflow with ARCNM?
Add the MCP server (https://api.arcnm.io/mcp-server/mcp) to the agent. The agent can start a calculation itself. It gets an upload link for a person to drop the STEP or PDF file on. Or it uploads the file directly. Then it starts the pricing. Through typed tools, it can also read the itemised result, explain the cost and rank the cost drivers. It can optimise the lot size and compare calculations or environments. Billing, credentials and tenant administration stay with people and are deliberately not exposed to agents.
How do I get my ERP cost history into the calibration?
Export part numbers with their real costs. Run the calibration preflight to see the coverage. Then post the set to the auto-calibrate endpoint of the costing environment. From then on every calculation in that environment carries your price level, with a confidence band.
Is the extracted part data mine to reuse?
Yes. Every calculation returns structured, versioned part data. It covers features, dimensions, tolerances, GD&T, treatments and materials. Reuse it for master-data management, similarity search or to fill PLM master data. We never train on it.