What is MCP (the Model Context Protocol)?
The Model Context Protocol is an open protocol for connecting language-model applications to external tools, data sources and prompts through a common interface. Instead of every application writing bespoke integrations for every system, a server exposes its capabilities once over MCP and any compatible client can use them. It standardises the plumbing between models and tools; it does not decide whether a model uses a tool well.
An MCP server advertises tools (callable functions), resources (readable data) and prompts. A client — a chat application, an IDE or an agent runtime — connects over a defined transport, discovers what the server offers, and passes tool definitions to the model. When the model emits a tool call the client routes it to the server, returns the result, and the loop continues. Authentication and authorisation remain the host application''s responsibility.
A company runs one MCP server in front of its ticketing system. The same server then serves its internal agent, its developer IDE assistant and its support console, with a single audit trail.
MCP reduces integration cost and vendor lock-in at the connection layer, which matters as organisations run several model providers. It also concentrates risk: a single server with broad permissions becomes a high-value target, so scoping and logging matter more, not less.
- That MCP makes agents reliable. It standardises access; step-level failure modes are unchanged.
- That MCP handles permissions. The host application still enforces who may call what.
- The specification is evolving; details of transports and auth have changed across revisions.
- Adoption varies widely by client, so capability support is uneven in practice.