Pick your systems, data sources and interfaces. We calculate setup, hosting and processing separately and draw the system architecture as you go. The same inputs always give the same result.
The cost of an enterprise setup depends mostly on infrastructure, compliance level and integration depth. Instead of a calculator we start with a 30-minute call and fix the scope in writing.
Edit the drawing as well: + Add a part, or remove one with the × on its corner. The price follows.
Prices are kept in euros; dollar amounts use a fixed exchange rate. The table is published with a version number and date and cannot be changed by visitors. What binds is our written quote.
A Fast deployment starts at €250 per working day. The base setup covers scoping, the first workflow and a test run; every connector, data source and interface adds its own days, shown next to it above. Hosting and processing are monthly and shown separately, so you see the one-off and the running cost apart.
n8n is the default: visual workflows that run on your server or on an EU server we manage. Make suits teams that already pay for it and want no server. LangGraph is for agents that need state, loops and human approval in code, and takes a little longer to build. A custom service is for when none of these fit.
A managed EU server sized to your load: the orchestration layer, the vector database for the knowledge base and the workers that process documents. The estimator picks the smallest server that holds your archive and users, and tells you which one.
Most of the work is in the document, not the model call. A simple e-mail or form costs a few cents; a scanned contract with tables needs a vision model and more checking. You set the share of complex documents, and the estimator also shows the model cost inside the price.
No. It is a deterministic estimate from published rates, meant to show what drives the cost. Only a written quote from LMXAI is binding. Send your setup from this page and we reply with one within a working day.
Agents that keep working outside office hours: webhook loops, scheduled jobs and multi-step agent graphs that run through the night and leave the results for the morning.
When the system must run on your own GPUs, falls under the high-risk rules of the EU AI Act, needs SSO and an SLA, or grows beyond the largest server here. The estimator says so when your inputs get there.