In this order of preference, and never the other way round.
The route is chosen by what the system actually supports, not by what is quickest for us. A system with no modern API is a normal Tuesday, not a reason to say no.
API
Official, documented, versioned. Where it exists this is always the answer.
Database
Read directly, write through a controlled layer with the vendor's blessing where the vendor still exists.
Files
Scheduled exports and imports. Unglamorous, entirely reliable, and often what a twenty-year-old system was designed for.
The interface itself
UI automation, when there is genuinely nothing else. It is the last resort and we will say so out loud rather than quietly billing for it.
60+ integrated, 44 documented. Both are true.
We have experience with more than sixty systems; 44 of them can be named publicly and are listed below.
The ones we can name.
Not on the list does not mean not possible. It usually means we have connected it and not written it down yet.
ERP · 11
SAP · Microsoft Dynamics 365 · Business Central · Helios · K2 · Abra · Money S5 · Pohoda · Oracle NetSuite · Byznys · Karat
CRM · 7
Salesforce · HubSpot · Dynamics CRM · Pipedrive · Raynet · Zoho · Anabix
E-commerce · 6
Shoptet · Shopify · WooCommerce · PrestaShop · Upgates · Magento
Communication and data · 8
Microsoft 365 · Teams · Outlook · Slack · WhatsApp Business · SharePoint · Google Workspace · Power BI
Automation · 5
n8n · Make · Zapier · Power Automate · Power Apps
AI models · 7
OpenAI / GPT · Claude (Anthropic) · Google Gemini · Azure OpenAI · Mistral · Llama · Microsoft Copilot
The same three rules on every integration we build.
A human approves anything sensitive.
Defined per step, named in the design document, and enforced in the code rather than in a policy nobody reads.
Everything lands in an audit trail.
Who or what did it, when, on which record, and what the previous value was. Written so an auditor can follow it without us in the room.
Roles and monitoring on day one.
Not retrofitted after go-live, when it is both more expensive and less complete.
European clients
Hosting and data residency follow your architecture.
Your own tenant, your own servers, or an EU cloud region. Hosting regions, subprocessors, retention and access controls are agreed before implementation and documented in the architecture and DPA. We design solutions to support your GDPR and EU AI Act requirements.
US partners
We start with a conversation, not a data-centre map.
With American partners the first work is consulting: understanding what you already run, where your own delivery is stretched, and where we can genuinely help. Hosting follows from that decision rather than leading it.
Talk it through →
Some of it runs in the EU cloud. Some of it runs on a machine you can point at.
Where it runs is your decision, made before we build. For sensitive data that is usually your own infrastructure; for everything else, EU cloud. Either way we are the ones operating it.
Tell us what you run. We will tell you what is reachable.
Book 30 minutes.
Bring a list of your systems. We will be straight about which ones are easy, which are work, and which are not worth it.
Book a call →
See how we build agents.
What sits on top of these connections once they exist.
How we build agents →
Or start with a routine.
Eight processes we have solved often enough to quote quickly.
All eight routines →