The Model Context Protocol (MCP) is an open standard for connecting AI agents to external data sources, tools, and services through a unified interface.
Think of MCP as a standard plug for agent integrations — instead of building custom connectors for every database, API, or file system, tools expose an MCP server that any compatible agent runtime can connect to.
What MCP provides
- Tools — Callable functions the agent can invoke
- Resources — Readable data sources (files, database records, API responses)
- Prompts — Pre-built prompt templates for common tasks
Why MCP matters
Before MCP, every agent framework had its own tool integration pattern. MCP standardizes the connection layer so:
- Tool builders write one MCP server, agents everywhere can use it
- Agent runtimes swap tools without rewriting integration code
- Security and permissions are handled at the protocol level
MCP in the agent stack
MCP sits between the agent runtime and external systems:
Agent (LLM + loop) → MCP client → MCP server → Database / API / Files
Frameworks like Eve, Claude Desktop, and others support MCP natively. It complements (rather than replaces) function calling — MCP is the transport; function calling is how the model selects what to invoke.
Related reading
See What Is Eve? for how MCP fits into Vercel's agent framework, and Tool Use for the broader pattern.