What is Model Context Protocol?
Also known as: MCP
Model Context Protocol (MCP) is a standardized way for AI agents to structure, exchange, and manage contextual information across model calls and tools.
MCP defines a lightweight message format and set of rules that let agents package relevant history, tool results, user state, and instructions into a consistent context object that any compliant model can read.
It separates transient working memory from long-term memory, supports selective compression, and includes versioning so agents can resume or hand off tasks without losing critical details.
By providing a common schema, MCP reduces prompt-engineering hacks and makes it easier to swap models or compose multi-agent systems while preserving shared understanding.
Example
An MCP-enabled travel agent packages the user's destination, budget, past flight preferences, and the result of a hotel search into one context packet; the next model call receives exactly that packet and can continue planning without re-reading the entire chat history.
Why it matters
As agents grow more complex and multi-model, a shared context protocol reduces errors, improves interoperability, and makes long-running tasks reliable.
Frequently asked questions
No, it is an emerging community proposal rather than a ratified standard like HTTP.
Related terms
A context window is the maximum number of tokens an LLM can process together in one pass, including the user's input and any conversation history.
ReAct is an agent prompting technique that interleaves explicit reasoning steps with tool-using actions so an LLM can dynamically plan, act, and observe results in a loop.
Tool Use (aka Function Calling) lets AI agents call external tools, APIs, or functions by outputting structured requests instead of just text.
In AI agents, memory refers to the component that stores and retrieves information from past interactions or experiences, enabling the agent to maintain context and make better decisions over time.
Agentic AI refers to AI systems that function as autonomous agents, capable of setting goals, making decisions, and taking actions to complete tasks with limited human input.
Agent memory is the component in AI agents that stores and retrieves information from past interactions, enabling recall of context, facts, or experiences to inform future actions.