Remember the drawer full of charging cables that a single connector standard made obsolete? That is pretty much the feeling we had when we implemented an AI integration via MCP for the first time.

One protocol instead of a hundred special cases

The Model Context Protocol, introduced by Anthropic in late 2024 and adopted across the industry in 2025 — by now practically all major providers support it — standardises the connection between AI models and the outside world: databases, file stores, CRM systems, internal tools. Before MCP, every combination of model and system needed its own hand-knitted adapter. With MCP, a system describes once which tools and data it offers — and any compatible AI model can use them. Connect once, use everywhere: hence the nickname “USB-C for AI”.

What this changes in practice

For companies, the question shifts from “Can we connect this?” to “What do we want to release?”. An MCP server in front of the ERP system can, for example, offer read-only access to inventory levels — no more, no less. This makes integrations not only faster but also more controllable: the permissions are explicit, auditable and centrally maintainable. In our projects this has drastically reduced the integration effort for new AI use cases — what used to take weeks is now often a matter of days.

The strategic view

Standards unfold their value through adoption, and that is exactly what is happening right now. Anyone who makes internal systems MCP-capable today builds a foundation on which future AI applications can simply dock — regardless of which model leads in two years. That is the real point: MCP is an investment in freedom to switch.

Which of your systems are worth connecting via MCP first can usually be answered quickly — just ask us.