MCP (Model Context Protocol) is an open communication protocol that solves a fundamental AI problem: how to connect a language model to external data sources, software tools, and third-party services in a standardized way. Created by Anthropic in November 2024, it defines a universal interface that lets AI read files, run functions, and receive contextual instructions — all through a single protocol built on JSON-RPC 2.0.
MCP adoption has been remarkably fast: OpenAI, Google, and Microsoft integrated it into their ecosystems in 2025, and the protocol was transferred to the Agentic AI Foundation under the Linux Foundation to guarantee open governance. More than 13,000 MCP servers are now available on GitHub, covering integrations with databases, document management systems, legal APIs, and many other services.
For the legal sector, MCP is a defining advance. It makes it possible to connect an LLM directly to legal databases, matter management systems, billing tools, or e-signature platforms. This standardized interoperability is the technological building block that makes legal AI agents possible: agents capable of acting end to end within the digital ecosystem of a law firm or a legal department.