MCP
Model Context Protocol
A standard way to plug external tool servers into an agent harness.
The Model Context Protocol is a standard for exposing external tool servers to a harness. It is how an agent gets tools beyond what the harness ships with: the agent never calls MCP directly, it calls a tool that the harness obtained from an MCP server.
The protocol solves a real integration problem. Without a standard, every harness would need its own integration for every system. With one, the integration is written once as a server, and any compatible harness can use it.
The cost is paid in context. Every advertised tool arrives as a definition, name, description, and parameter schema, and the model can only call tools it knows about. Install several generous servers and a session starts with thousands of tokens of tool definitions before you type anything. Tool search, where the context holds a pointer instead of the full definitions, is the current mitigation.
Related terms
Tool
A function the model is allowed to call, like reading a file or running a command.
Harness
Everything around the model that turns it into an agent: tools, system prompt, permissions, hooks.
Context Engineering
Deciding what goes into the model's context, and what stays out.
Integration
Connecting your system to the other systems it needs to work with.
Agent
A model with tools, a system prompt, and a loop that lets it act and observe results.
Want this working inside your team?
It starts with one conversation about your workflow.