Anthropic announced the Model Context Protocol on 25 November 2024 as an open standard for connecting AI assistants to the systems where data lives. It standardises how applications expose tools, resources and prompts to language models. Instead of bespoke glue for every model–tool pair, a server implements MCP once and any MCP-aware client can use it — the official documentation compares it to a "USB-C port" for AI applications.
In December 2025 Anthropic donated MCP to the newly formed Agentic AI Foundation, a directed fund under the Linux Foundation co-founded by Anthropic, Block and OpenAI, giving the protocol vendor-neutral governance. Its existing maintainers still run day-to-day development. MCP matters as assistants become agentic: a shared protocol makes connectors composable and portable across vendors.
What is MCP and how does it work?
MCP uses a client–server architecture with three roles:
| Role | What it is | Example |
|---|---|---|
| Host | The AI application the user works in | A chat app, an IDE such as Visual Studio Code, or an agent |
| Client | A connector inside the host; one per server | The link VS Code opens to each server it uses |
| Server | A program that exposes a system to the model | A filesystem, database, calendar or code-repository server |
Servers offer three core primitives: tools the model can call to take actions, resources that supply context such as file contents or records, and prompts, reusable templates for common interactions. Messages are exchanged as JSON-RPC 2.0, either locally over standard input/output or remotely over Streamable HTTP, where OAuth is the recommended way to authorise access.
Why the Model Context Protocol matters
Before MCP, every assistant needed its own one-off integration for every data source. With a common protocol, a connector written once works in any compatible client; the MCP documentation lists Claude, ChatGPT, Visual Studio Code and Cursor among the applications that support it. MCP builds on the same idea as tool use / function calling, but standardises it across vendors.