Introduced by Anthropic in late 2024 and now adopted across the AI ecosystem, MCP is often described as a universal connector for AI: it standardises how a model discovers a tool, learns its inputs, calls it and reads the result. Without it, every assistant-to-system link is a one-off integration; with it, any MCP-aware agent can use any MCP server.
For Digital Product Passports this turns compliance into something an agent can do, not just a human clicking a UI. The TracePass MCP server exposes passport operations — search products, draft a passport from documents, check confidence scores, publish a GS1 Digital Link QR — so an AI workflow can run the regulated process end to end with a human approving before publication.
Frequently asked
Who created MCP and is it open?
MCP was introduced by Anthropic in late 2024 as an open protocol and has since been adopted across the wider AI ecosystem. Because it is open, any model or agent platform can implement a client, and any service can expose an MCP server — no single vendor controls it.
What can an AI agent do with the TracePass MCP server?
It can run the passport workflow programmatically — find products, draft a Digital Product Passport from uploaded documents, read each field's confidence score and source attribution, and publish a GS1 Digital Link QR. A human still approves the regulated data before it goes live.