Model Context Protocol

An open protocol by Anthropic that standardizes how AI applications provide context — tools, resources, and prompts — to language models.

What Is the Model Context Protocol?

The Model Context Protocol (MCP) is an open standard introduced by Anthropic in late 2024 that defines a universal interface for connecting AI models to external data sources, tools, and capabilities. Its goal is to solve a fragmentation problem: every AI application was building its own bespoke integration layer, making it impossible to share tools across applications or models.

MCP provides a common language so that a tool built once can be used by any MCP-compatible AI application — whether that is Claude, a custom agent, or a development environment like Cursor or VS Code.

The Problem MCP Solves

Before MCP, if you wanted your AI assistant to access a database, a developer built a custom integration for that specific assistant. If you wanted a different assistant to use the same database, you built another custom integration. Multiply this across thousands of tools and dozens of applications, and you get an unmaintainable web of one-off connectors.

MCP defines a standard so that:

Tools are built once and work everywhere.
Applications do not need to know the implementation details of each tool.
Users can compose capabilities from multiple MCP servers without developer intervention.

Core Concepts in MCP

MCP is built around three primitives:

Tools — Functions the model can invoke (equivalent to function calling). For example, a KnowledgeSDK MCP server might expose extract_url, scrape_url, and search_knowledge as tools.
Resources — Data sources the model can read, like files, database records, or API responses.
Prompts — Reusable prompt templates that can be parameterized and served to the model on demand.

How MCP Works

An MCP server exposes capabilities over a standard transport (stdio or HTTP with Server-Sent Events). An MCP client (the AI application or agent) connects to one or more servers, discovers their available tools and resources, and can call them during a conversation.

The workflow:

Agent connects to an MCP server on startup.
Agent queries the server for available tools (tools/list).
During a task, the agent calls a tool (tools/call) with arguments.
The server executes and returns a result.
The agent incorporates the result into its reasoning.

KnowledgeSDK and MCP

KnowledgeSDK publishes @knowledgesdk/mcp, an MCP server that exposes its web intelligence capabilities — extraction, scraping, screenshot, classification, sitemap, and search — as standard MCP tools. Any MCP-compatible agent or IDE can add KnowledgeSDK capabilities without writing custom integration code.

Why MCP Matters for the Ecosystem

MCP shifts the model for AI tooling from a hub-and-spoke architecture (each application integrates each tool separately) to a mesh architecture (tools and applications speak a common protocol). This creates a marketplace dynamic: tool builders can publish MCP servers, and application builders can compose them freely — accelerating the entire ecosystem.