What Is an MCP Registry?

📘 What a catalogue entry contains

A registry, in this vocabulary, is a catalogue of MCP servers. An entry names a server, describes what it is for, and records how to reach it: what to launch or what address to connect to, and what configuration the connection needs. It is a directory rather than a runtime component. Nothing in the protocol requires one, and a connection works identically whether the details came from a catalogue or from a configuration file typed by hand.

What a catalogue settles is the part that happens before any connection exists — which server, obtained from where, reached how. What it does not settle is the part that happens after. What a server actually publishes is learned from the server itself, by listing its operations, resources and prompts once the session is open. A catalogue that also lists capabilities is describing a snapshot; the server remains the only accurate account of its own surface.

An entry is a claim by whoever wrote it, and this is the part worth being blunt about. A catalogue can record that a server exists, who published it, and how to run it. It cannot establish that the code does what the entry says, that the publisher is who they say they are, or that the artefact fetched today is the one somebody reviewed last month. Choosing a server from a catalogue is choosing to run a program.

A usable entry has to carry three things, and the third is the one most often absent. An identifier stable enough to name the same server twice. A means of obtaining it — a package, an image, an address — which is the point at which an entry stops describing and starts instructing. And the configuration a connection needs before it will work at all: what to launch it with, what environment values it reads, what credential it expects to be given. An entry that names a server and omits the third has not saved anyone the work, it has moved the work to a support thread.

🧩 When the catalogue and the server disagree

The server is right. An entry describes what somebody observed or intended on the day they wrote it; a running server answers for itself at the moment it is asked, and that answer is the only account current by construction. An integration built against a capability list copied out of a catalogue is built against a snapshot nothing keeps in step.

This is also why a version belongs in an entry. A bare name points at whatever that name resolves to today, which may be a different program from the one somebody read, on a schedule set by somebody else. Recording what was reviewed and recording what is being run is the discipline any dependency gets, and appearing in a catalogue exempts nothing from it.

📐 An entry is a claim, not a guarantee

This page describes what a catalogue is and what it settles. It names none, ranks none and recommends none: a list of third-party servers is a large number of factual assertions with a short shelf life, published on a page that carries a price.

⚙️ How the Registry classifies searching a catalogue

Searching a catalogue for a server that fits and recording its connection details is retrieval over a body of entries that already exists, which the information retrieval class covers. On the order form's task class field, select Information retrieval, which covers search, information gathering, research, summarisation. That class has its own page at /tasks/retrieval.

The Registry files this topic under one canonical task, fixed in advance and published here rather than generated per visit:

Search a catalogue of MCP servers for one that exposes a calendar service and record its connection details.

That string, hashed once, is a32a9343436110c5… — the first sixteen of sixty-four hex characters. An assessment ordered against a task of this kind is computed from the class, the wording submitted, and the permanent chart the Registry derives for the model configuration named on the order: its ascendant, its ruling planet, and the composite harmony that constrains the range of verdicts available to it. The whole derivation is published at /method.

The risk in catalogue work is provenance: an entry that resolves to a different artefact than the one that was read, or a description written by somebody with no stake in its accuracy. 4 risk factors are tracked for information retrieval tasks generally; an assessment reports 3 of them, selected by the submitted task's own seed rather than chosen. Their names, severities and descriptions are certificate content and are not published on a free page.

Get Assessment

📏 What this page does not claim

The Registry makes no representation that an assessment predicts the outcome of any task, and no assessment is validated against real-world results.

This page defines a term and states a classification. It publishes no verdict: the Registry does not test implementations, does not measure how often anything works, and has never run the canonical task above against any model. Which catalogue is consulted, and what would make an entry acceptable to a particular operator, are decisions the canonical brief leaves open.

What is sold is a permanent, numbered assessment for one submitted task at one model configuration — a verdict on a seven-point scale, a recommended execution window, and, above the first tier, the risk factors the class tracks. EUR 1.90 Standard, EUR 4.90 Extended, and EUR 14.90 Full Chart, which adds the permanent chart for the configuration itself. Assessments from EUR 1.90; machine-readable at /pricing.json.

🔗 Related surfaces

← All MCP topics · The AFR-1 method · Information retrieval task class

Related topics: Listing and discovery · MCP server · Security considerations

❓ Questions about a registry

What is an MCP registry?

A catalogue of MCP servers and their connection details: what each server is for, how to reach or launch it, and what configuration it needs. It is a directory consulted before a connection is opened, not a part of the protocol.

Is a registry required to use MCP?

No. A host connects to a server using configuration it holds, however that configuration was obtained. A catalogue is a convenience for finding servers and their details, not a requirement of the protocol.

Does a catalogue entry mean a server is safe to run?

No. An entry records what its author wrote. It cannot establish that the code matches the description, or that today’s artefact is the one somebody reviewed. Selecting a server is selecting a program to run with whatever access it is given.

Can a catalogue tell me what a server exposes?

It can record what somebody wrote down. The current answer comes from the server: once a connection is open, a client asks what is published and receives the names, descriptions and argument schemas as they are now. Where the two differ, the server is the account that is current.