MCP (Model Context Protocol)
In one sentence MCP is an open standard that lets any AI assistant connect to any tool or data source through one common protocol, instead of custom integrations for each pair.
Updated
The Model Context Protocol, or MCP, is an open standard for connecting AI applications to external tools and data sources — one protocol replacing a tangle of custom integrations.
It is USB for AI tools. Before USB, every device needed its own port and cable; every printer-computer pair was its own engineering project. One standard connector ended that: any device, any machine. Before MCP, connecting an assistant to Slack, a database and GitHub meant custom function-calling glue for each pair of assistant and service — the M×N problem. With MCP, a service implements one server, an AI app implements one client, and any client can use any server: M+N.
The mechanics: an MCP server is a small program exposing three kinds of things — tools (actions the model may request: query this database, create this ticket), resources (data it can read: files, schemas), and prompts (reusable templates). The AI application connects, lists what is available, and hands those tools to the model, which uses them through ordinary function calling. Servers run locally or remotely, speaking JSON-RPC (a simple remote-call format).
Released open-source by Anthropic in late 2024, it became the rare standard that actually took: OpenAI, Google and Microsoft adopted it, and thousands of community servers exist for databases, browsers, and SaaS tools. Practical notes: an assistant with MCP tools is an agent with real reach — so permissions matter — and a malicious or compromised server's tool descriptions are a prompt-injection surface. Grant access accordingly.
Where to go next
- Full lesson: Function calling
- Related terms: function-calling, agent, rag, guardrails