Model Context Protocols (MCPs) emerge

Most company chatbots still break the same way: they can answer questions from one system, but the moment you need Salesforce, Slack, a database, and a calendar in one task, somebody has to build custom glue for every connection. Anthropic launched Model Context Protocol in November 2024 to replace that one-off wiring with one shared standard. (anthropic.com) A protocol is just a common language for software, like shipping containers gave ports one box size instead of a different crate for every cargo company. The Model Context Protocol gives artificial intelligence apps one way to ask for data and one way to call tools across many systems. (modelcontextprotocol.io) The basic setup has three parts. A host is the app a person uses, a client is the connector inside that app, and a server is the program that exposes a file system, database, Slack workspace, or other system to the model. (modelcontextprotocol.io) Those servers expose three different things. Tools let the model do an action, resources give it read-only context like files or schemas, and prompts provide reusable instruction templates for common workflows. (modelcontextprotocol.io) That sounds abstract until you compare it with the old way. Anthropic said every new data source used to require its own custom implementation, while later MCP explainers described the old integration mess as an N times M problem where every new model and every new system multiplied the number of connectors. (anthropic.com) (cloudsummit.eu) Model Context Protocol does not replace every application programming interface. It sits above those interfaces, so a company can keep its existing systems while exposing them through lightweight Model Context Protocol servers that an agent can discover and use at runtime. (modelcontextprotocol.io) (techcommunity.microsoft.com) That discovery piece is why people in enterprise software care. Instead of hardcoding every function in advance, the protocol lets an agent ask what tools exist, what inputs they take, and what data is available during a live session. (modelcontextprotocol.io) (cloudsummit.eu) The protocol also keeps state across a session, which is a fancy way of saying the agent can remember what it already pulled and what step it is on. The specification uses JSON-RPC 2.0 messages, supports capability negotiation, and includes utilities like progress tracking, cancellation, logging, and error reporting. (modelcontextprotocol.io) Security is the part that decides whether this becomes a real enterprise layer or a toy. The specification says users must explicitly consent to data access and actions, hosts must protect user data with access controls, and tool execution has to be treated as potentially dangerous because tools can trigger code or outside systems. (modelcontextprotocol.io) The story moved from interesting idea to industry signal on December 9, 2025, when Anthropic donated Model Context Protocol to the Agentic AI Foundation under the Linux Foundation. Anthropic said the foundation was co-founded by Anthropic, Block, and OpenAI, with support from Google, Microsoft, Amazon Web Services, Cloudflare, and Bloomberg. (anthropic.com) By that December 2025 update, Anthropic said there were more than 10,000 active public Model Context Protocol servers, adoption across ChatGPT, Cursor, Gemini, Microsoft Copilot, and Visual Studio Code, and more than 75 connectors in Claude’s directory. The same announcement said official software development kits were seeing more than 97 million monthly downloads across Python and TypeScript. (anthropic.com) So the news is not that companies suddenly stopped using application programming interfaces. The news is that a governed middle layer is emerging between models and business systems, and if Model Context Protocol keeps spreading, enterprises may wire agents into workflows the way developers once standardized on web browsers, USB-C ports, and the Language Server Protocol. (modelcontextprotocol.io) (anthropic.com)