USB for AI - a standard way for agents to plug into any tool or data source.
Before USB, every device had its own connector - a different cable for every printer, camera, and keyboard. MCP is solving the same problem for AI agents. Instead of building custom integrations for every tool an agent might need, MCP provides one standardized protocol that any tool can implement and any agent can use.
MCP (Model Context Protocol) is an open protocol developed by Anthropic that lets AI models connect to external data sources and tools. An MCP server exposes capabilities - like reading from a database, searching a knowledge base, or calling an API - in a format that any MCP-compatible client (like Claude Code) can automatically discover and use.
The protocol is straightforward: an MCP server describes what tools it offers and what inputs they expect, and the AI client can then call those tools as part of its workflow. No custom integrations, no glue code, no middleware.
The power of a coding agent is directly proportional to the tools it can access. An agent that can only read and write files is limited. An agent that can also search documentation, query databases, check CI status, read Slack messages, and browse internal wikis can handle dramatically more complex tasks.
MCP makes this extensible. Instead of waiting for your AI provider to build an integration with every tool you use, you can run (or build) an MCP server for any tool and instantly give your agent access to it. The ecosystem is already large and growing fast.