MCP (Model Context Protocol)
An open standard that lets AI applications connect to external tools, data sources, and services through a common interface, so any compliant agent can use any compliant tool without custom integration work.
Also known as: Model Context Protocol, MCP server
The Model Context Protocol is an open standard for connecting AI models to the outside world. Introduced by Anthropic in late 2024 and adopted widely across the agent ecosystem, it defines a common way for an AI application to discover and call external capabilities: reading files, querying databases, hitting APIs, or invoking specialised services. An MCP server exposes a set of tools and resources; an MCP-aware client, such as a coding agent or a chat application, connects to that server and can then use those tools without any integration code written specifically for it.
The value is composability. Before a shared protocol, every tool needed a custom integration for every agent that wanted to use it, an N-times-M problem that kept useful capabilities siloed. With MCP, a tool built once works with every compliant client, and an agent gains access to a growing catalogue of servers by speaking one protocol. This is why so many developer tools now ship an MCP server as a first-class interface.
For DeAI, MCP matters because it is the wiring through which decentralised services reach agents. A decentralised storage or memory network can expose itself as an MCP server, and any MCP-aware agent can then store and recall data through it without knowing anything about the underlying blockchain. Walrus Memory, for instance, offers an MCP server so that agents in tools like Claude Code or Cursor can persist wallet-owned memory through the same standard interface they use for everything else. The protocol is becoming the connective tissue between agents and the services, centralised or not, that they depend on.