MCP (MODEL CONTEXT PROTOCOL)

An open standard for connecting AI applications to external tools and data sources through a common interface.

Zero retention: everything you type stays in this browser. Verify it yourself in DevTools.
Plain-language definition

What Does MCP (Model Context Protocol) Mean?

MCP, the Model Context Protocol, is an open standard for connecting AI applications to tools and data. Anthropic introduced it in 2024, and it is now stewarded by the Agentic AI Foundation under the Linux Foundation. Instead of every AI app writing its own integration for every service, a service is wrapped once as an MCP server, which describes the tools it offers, such as "search issues" or "query database." Any MCP-capable client, such as a coding agent or a desktop chat app, can then connect to it and let the model call those tools. It is plumbing: it standardizes how tools are described and called, not what they do.

Why Does MCP (Model Context Protocol) Matter?

MCP is why you can plug the same database, issue tracker or browser tool into several different AI tools without custom code for each. It also widens what an agent can reach, which raises security questions: an MCP server runs with whatever access you give it, and text it returns can contain prompt injection. Install servers you trust and grant them only the permissions they need.

What Does MCP (Model Context Protocol) Look Like in Practice?

A team adds a GitHub MCP server to their coding agent. Now, when asked "summarize open bugs labeled checkout," the agent calls the server list_issues tool, gets back structured results, and writes the summary, without anyone copying issues into the chat. The same server works in other MCP clients they use. They give it a read-only token, so even a confused or injected request cannot close or edit issues.

What Do People Get Wrong About MCP (Model Context Protocol)?

MCP does not make a model smarter or safer by itself. It only standardizes the connection. A badly designed or untrusted MCP server can still return wrong data or injected instructions.

Which Terms Are Related?

  • Tool use: A model asking its application to run a defined function, such as search or run tests, and using the result.
  • Agentic workflow: A setup where an AI takes multiple steps on its own, using tools and checking results, rather than giving one reply.
  • Prompt injection: An attack where instructions hidden in content an AI reads trick it into acting against its user.
  • Guardrail: A rule or check, in the instructions or in code, that keeps an AI system from doing what it should not.
  • Context engineering: Deciding what goes into a model context window, including instructions, files and tool output, so the model can do the task well.

All 28 terms are in the AI glossary. For models side by side, see frontier vs open-weight.

Frequently Asked Questions

Is MCP only for Claude?

No. Anthropic created it, but it is an open standard now stewarded by the Agentic AI Foundation, and many AI clients and agents from different vendors support it.

Is MCP the same as an API?

An MCP server usually wraps an existing API. MCP adds a common way for AI clients to discover the available tools, see their descriptions and call them, so one integration works across many clients.

Are MCP servers safe to install?

Treat them like any software you install. A server can read or change whatever its credentials allow, and its output goes straight into the model context. Prefer trusted sources and least-privilege tokens.

Was this helpful?