Flashcard: why connect agents through MCP?

Attempt it

MCP is an open protocol, introduced by Anthropic in late 2024 and widely adopted across 2025-2026, that standardizes how an LLM agent connects to tools, data sources, and prompt providers. An MCP server exposes capabilities (tools to call, resources to read, prompts to use) over a stable interface, and any MCP-compatible client (Claude, Cursor, agent runtimes like LangGraph or CrewAI) can use them without bespoke integration code. The point is reducing N-by-M integration to N-plus-M: instead of every agent framework defining its own tool format and every tool provider building one adapter per framework, both sides target MCP and the cross-product is gone. Agents use MCP because it lets them gain new capabilities by pointing at a server URL rather than writing custom tool wrappers, and it lets tool providers reach every MCP-compatible client by building once. By 2026, MCP servers exist for filesystems, databases, GitHub, Slack, browser automation, and most major SaaS products.

Your answer

**MCP** stands for Model Context Protocol. It is an open standard introduced by Anthropic in late 2024 for how an LLM agent connects to tools, data sources, and prompts. By 2026 it is widely adopted across the agent ecosystem. An MCP server exposes three kinds of capabilities: tools the model can call, resources the model can read, and prompts the model can use. Any MCP-compatible client (Claude desktop, Cursor, LangGraph) can connect to that server and use those capabilities without writing custom adapter code. Why this matters. Before MCP, every agent framework had its own tool format. With N frameworks and M tools, the world needed N-times-M custom integrations. MCP collapses this: a tool provider writes one MCP server and every client can use it; a framework writes one MCP client and the agent can use every MCP tool. Total work drops to N-plus-M. Agents use MCP because it lets them gain new capabilities by pointing at a server URL rather than writing bespoke tool wrappers. In 2026, MCP servers exist for filesystems, databases, GitHub, Slack, browser automation, and most major SaaS products.

**MCP** is the Model Context Protocol, an open standard introduced by Anthropic in late 2024 that standardizes how an LLM agent connects to tools, data sources, and prompt providers. By 2026 it is widely adopted: Claude desktop, Cursor, OpenAI Agents SDK, LangGraph, CrewAI, and AutoGen all speak MCP natively or have first-class support. An MCP server exposes three capability primitives. **Tools** are functions the model can call. **Resources** are data the model can read (files, database rows, web pages). **Prompts** are reusable prompt templates the user or agent can invoke. The server runs as a separate process (stdio or HTTP) and any MCP-compatible client can connect. The motivation is reducing N-times-M integration work to N-plus-M. Before MCP, every framework defined its own tool format and every tool provider built one adapter per framework. With MCP, frameworks target the protocol once as clients and tools target it once as servers; total work is N-plus-M. Agents use MCP because it lets them add capabilities by pointing at a server URL rather than writing custom wrappers. Drop in a GitHub MCP server and the agent has GitHub access. Run a filesystem server locally and the agent can read and edit files. The ecosystem effect is the real win. By 2026, public MCP servers exist for filesystems, Postgres, GitHub, Slack, Linear, Notion, Stripe, browser automation, and most major SaaS products. A new framework that speaks MCP gets that entire ecosystem for free.

Quick reference

TL;DR

MCP is an open protocol from Anthropic (late 2024) that standardizes how LLM agents connect to tools, data, and prompts. It collapses N-by-M custom integrations into N-plus-M shared interfaces.

Memory aid

Easy to grasp

Imagine every brand of phone needing its own special charger, so a Samsung charger does not work on an iPhone and an iPhone charger does not work on a Pixel. That is the world before MCP: every agent framework had its own way to talk to tools, so every tool had to build a separate adapter for every framework. MCP is like USB-C for AI agents: one standard plug. A tool builder makes one MCP server, and any MCP-compatible client (Claude, Cursor, LangGraph) can use it. An agent builder makes one MCP client, and the agent can use any MCP-compatible tool. The number of integrations to build drops from N-times-M to N-plus-M, which is why MCP spread fast across 2025-2026.

Key concepts

Model Context Protocol Tool Protocol Interoperability Function Calling

Detailed answer & concept explanation

~7 min read

Everything you need to truly understand this topic: intuition, mechanics, step by step explanation, code, formulas, and worked example. Click to expand.

Everything you need to truly understand this topic: intuition, mechanics, step by step explanation, code, formulas, and worked example.

Quick scan · 2–4 min

Everything important, quickly.

Define MCP as an open protocol from Anthropic, late 2024, for connecting LLM agents to tools, data, and prompts. Name the three primitives: tools, resources, prompts. State the N-by-M to N-plus-M economic argument. Distinguish MCP from provider-native function calling. Close with the 2026 ecosystem breadth and the reasons the protocol spread quickly.

Real-world examples

Real products, models, and research that use this idea.

Anthropic's Claude desktop app ships with MCP support out of the box; users add a filesystem or GitHub MCP server and Claude immediately gains those capabilities.
Cursor IDE uses MCP to expose its codebase tools to the agent, and lets users add third-party MCP servers for custom integrations.
OpenAI announced first-class MCP support in the Agents SDK in 2025, allowing GPT-5.5 agents to use the same MCP servers as Claude agents.
Public MCP server registries (mcp-servers community catalogs) list hundreds of servers for filesystems, databases, GitHub, Slack, Linear, Notion, Stripe, and most major SaaS products.

Interviewer follow-ups

What an interviewer would ask next. Try answering before peeking at the approach.

QHow is MCP different from provider-native function calling?

QWhy did MCP take off so quickly across 2025?

QWhat are the three capability types an MCP server can expose, and how do they differ?

Don't say this

Red flags and common mistakes that signal junior thinking. Click to expand.

Most common mistake

Calling MCP an Anthropic-only protocol. It was introduced by Anthropic but is open and now used across Claude, Cursor, OpenAI Agents SDK, and most agent frameworks. The open-protocol property is what made it spread.

60-second cheat sheet

The night-before-the-interview bullets. Scan these on the way to the call.

References

Primary sources. Skim if you want the original framing.