Free tools Windows power users keep installed
One-click scans. No signup required.
Model Context Protocol (MCP) is a standard way for an AI application to discover and use capabilities offered by an external system. The AI application remains the place where the model experience lives; an MCP server offers data or operations, and an MCP client handles the protocol communication between them. MCP is not itself an AI model, database, or standalone agent. The official TypeScript SDK describes it as an open standard connecting AI applications to the systems where data and tools live.
How does MCP connect an AI application to an external system?
Think of MCP as a shared connector format. As an analogy, it is like a standard plug: services still own their data and decide what operations to make available, while a client uses a common interface to find and request those capabilities. That analogy is only a mental model; MCP itself is a protocol.
- Host: The application that contains the model experience—for example, an AI assistant.
- Client: The component in the host that manages protocol communication with an MCP server.
- Server: The component that offers capabilities or data to the client, such as tools, resources, or prompts.
The host and client do not automatically gain access to everything in an external system. They discover and request what a server makes available, and the server handles those requests according to its implementation.
MCP explained with an order-lookup example
The official TypeScript SDK calling guide illustrates the exchange with an order-support server. It is a documentation example, not evidence that a real merchant system was contacted.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
#1 Best Overall
- Discover: The client asks the server what tools it offers. The example server advertises a tool named
lookup-order. - Call: The client requests that tool by name, passing
{"id":"A-1041"}as its input. - Receive: The tool returns text saying order A-1041 has three items and has shipped. The AI application can then use that result in its response.
The same server can offer other kinds of capabilities. For example, the guide shows a resource identified as orders://recent, which the client can read for recent-order context. It also demonstrates obtaining prompts, including a user-selectable template for summarizing an order.
What are MCP tools, resources, and prompts?
MCP distinguishes these primitives by how they are used and controlled, as described in the specification overview. That overview is marked as a draft, so its status and revision may change.
| Primitive | How it is used | Example |
|---|---|---|
| Tool | A model-controlled call to an executable function. It can retrieve information or perform an action. | Look up order A-1041 or export a report. |
| Resource | Application-controlled contextual data supplied to the model. | Recent orders or the contents of a file. |
| Prompt | A user-controlled, selectable template or instruction. | A “summarize this order” prompt. |
In practice, the client can list tools before calling one, read a resource, or retrieve a prompt. The server defines which of these it supports; MCP does not require every server to offer all three.
What MCP does—and does not—provide
- It standardizes the connection pattern. Different AI applications and external systems can use a shared protocol to discover and request capabilities.
- It does not supply the underlying system. A server must still connect to or contain the relevant data and implement the operations it offers.
- It does not do the model’s reasoning. The host’s model interprets the conversation and results; MCP carries requests and responses between client and server.
- It does not guarantee universal feature support. What is available depends on the capabilities implemented by the particular host, client, and server.
What MCP can also do: interactive interfaces
MCP Apps are an extension rather than one of the three basic primitives. In an announcement dated January 26, 2026, the maintainers described an approach where a tool carries UI metadata and the server supplies a UI resource that the host can render in a sandboxed iframe. The MCP Apps quickstart demonstrates a tool returning server time alongside an HTML UI resource. That tutorial requires Node.js 20 or later and assumes familiarity with tools and resources; the extension is not necessary to understand a basic MCP connection.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →SDK versions and implementation choices
SDK version statements are specific to the documentation and can change. The official Python SDK documentation identifies v2 as its current stable line, while its v1 documentation describes the maintenance line and points readers to v2. The Python documentation lists stdio, Streamable HTTP, and SSE transports. The official TypeScript SDK v2 documentation identifies its stable release line as implementing the 2026-07-28 specification revision.
Those SDK statements do not mean every host or server supports every feature in a given specification revision. For an implementation, choose a language and transport based on the systems you need to connect, how the client and server will communicate, which primitives are required, and whether an interactive UI extension is needed. The available documentation does not establish a universally best language or transport.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
