Glossary
Model Context Protocol (MCP)
MCP is an open protocol that lets an AI agent connect to external tools and data sources through a standard interface, so a tool built once works with any compatible agent.
The problem it solves
Before a standard, every agent platform needed its own integration with every tool — an N-by-M problem where both numbers were growing. MCP makes it N plus M: a tool exposes an MCP server once, and any MCP-capable agent can use it.
Client and server
Most agent platforms are MCP clients — they consume servers. A platform can also be a server, exposing its own agents as tools that other systems call.
What it does not solve
Authentication still expires, endpoints still move, and consent still lapses. A standard interface makes connections easier to establish, not self-maintaining — keeping them alive is ongoing work.
Related terms
FAQ
What is MCP?
The Model Context Protocol, an open standard that lets an AI agent connect to external tools and data sources through a common interface. A tool exposed once as an MCP server works with any MCP-capable agent, turning an N-by-M integration problem into N plus M.
Does MCP make connections maintenance-free?
No. It standardises how a connection is established, not how it stays alive. Tokens still expire, OAuth grants still lapse and providers still move endpoints, so reconnection remains ongoing work.