Skip to content

What MCP is and why Anthropic created it

MCP کیا ہے اور Anthropic نے اسے کیوں بنایا

35 min read

Three ways to see it

  1. MCP stands for Model Context Protocol. It is an open standard that defines how an AI model talks to external systems. The model is the client. The external system is wrapped inside a small server. The two communicate using a small set of message types: list_tools (what can you do?), call_tool (do this thing now), list_resources (what can you read?), read_resource (give me that file). The protocol is opinionated about the shape of these messages but agnostic about what is on the other side. The other side can be a file system, a database, a web search, an FBR API, an internal HR system, or a tea-maker controlled by a microcontroller.

  2. Why Anthropic invented this rather than letting every team continue to write their own glue. The combinatorial maths is brutal. If there are N model providers and M tools, the world needs N times M custom integrations. With ten models and ten thousand tools you have a hundred thousand integrations, each of which decays the moment one side updates. MCP cuts this to N plus M. Each model speaks MCP once. Each tool wraps in MCP once. The matrix collapses into a list. This is the same insight that drove TCP/IP and USB and HTTP. It looks simple in retrospect; it is the kind of insight an industry only swallows when one big player commits.

  3. Pakistani context. Imagine you run an MCP server that exposes FBR's NTN lookup, NADRA's CNIC verification (where regulated), and your own ERP's purchase orders. Any model that speaks MCP, Claude Desktop, ChatGPT through its plugin layer, Cursor, your in-house agent, can now query all three through one connector. A new model arrives next year; it speaks MCP; it inherits all your tools on day one. An employee writes a new internal lookup; it gets a server wrapper; every model has it on day one. The protocol is the contract that lets your organisation accumulate AI capability instead of constantly rebuilding it.

Quick check

Quick check: what makes modern AI different from a rule-based program?

The why-tree

Why-tree level one: why a new protocol rather than reusing OpenAPI? Because OpenAPI describes what a single API offers. MCP describes the conversation an LLM has across multiple APIs while reasoning. The two complement each other; many MCP servers internally wrap OpenAPI services.

Try this with Claude

AI-edge prompt to try with Claude or ChatGPT: 'Explain Model Context Protocol to my board of directors, three of whom are over 60 and not technical. Use the USB-C analogy, give one Pakistani example involving FBR or NADRA, and finish with the single sentence I should say to convince them to fund a six-month MCP pilot.' Rate the answer on whether your actual board would buy it.

Sources

Sources and further reading. Anthropic, 'Introducing the Model Context Protocol' (anthropic.com/news/model-context-protocol), November 2024. The MCP specification at modelcontextprotocol.io. The official servers repository at github.com/modelcontextprotocol/servers. Anthropic Engineering blog on tool use (docs.anthropic.com/en/docs/build-with-claude/tool-use). USB-C historical analogy for context: usb.org.

Mark complete + next lesson →