MCP, explained: what it actually connects, and what it doesn't.
Model Context Protocol gives an AI agent a standard way to call into a business's own systems: read live data, run a search, trigger a defined action. It has nothing to do with payment or identity. That's a separate job, done by separate protocols.
5 min read
What MCP is
Of all the protocols in this hub, MCP is the one most likely to already be partly set up somewhere, and most likely to be mistaken for solving more than it does.
MCP is an open standard for connecting an agent to a specific system's data and actions. A business (or, more often, a platform acting on its behalf) runs a small server that exposes a defined set of things an agent can call: check availability for these dates, look up this menu, return the current price for this item. The agent's model calls those directly, mid-conversation, the same way it might call any other tool.
What it lets an agent do that a website can't
A website is built for a person scrolling and clicking. MCP gives an agent the same information in a shape it can reason over directly: structured fields rather than page layout, a defined response rather than something that has to be scraped and guessed at. That's the difference between an agent inferring a jacket is probably in stock and an agent knowing it is, checked a second ago.
What MCP explicitly does not do
MCP has no built-in concept of payment authorisation, and no way to prove the agent calling it is allowed to act on a real person's behalf. It will happily tell an agent a desk lamp is £34; it says nothing about whether that agent is cleared to buy it. MCP tells an agent how to ask. It doesn't say whether it should be allowed to.
No integration to build from scratch
A business doesn't need to build an MCP server from scratch to be reachable this way. Selfe exposes the right set of tools automatically from a business's existing systems, so an agent gets live, structured answers without the business maintaining a second integration alongside its website.
Does having MCP set up mean an agent can already buy from us?
No. It means an agent can see accurate, live information. Buying still needs identity, authorisation and a checkout step, covered elsewhere in this hub.
Do we need our own developers to build this?
Not with Selfe. The tools are generated from the business's existing systems rather than hand-built per integration.
The plumbing behind every agent purchase.
Every purchase an AI agent completes, whatever it's buying, passes through the same four jobs: proving who the agent is, proving what it's allowed to spend, connecting to the business's own systems, and completing a transaction the business can settle. The protocols in this hub each do one of those jobs. None of them does all four.
A2A, explained: how one agent proves what it's asking another to do.
A shopping AI agent rarely finishes the whole job itself. It hands specific tasks to other, more specialised agents along the way. Agent2Agent protocol is how one agent describes a task to another clearly enough that the second agent can trust it and act on it.
UCP and the checkout handoff: what changes when a purchase completes inside the conversation.
Every other protocol in this hub gets a purchase ready. UCP is the one that completes it: the point where a business hands back a confirmed order without sending the customer off to finish the job on a website.