What is MCP (Model Context Protocol)?
MCP (Model Context Protocol) is an open protocol developed by Anthropic that standardizes the way language models (LLMs) connect with external data sources, tools, and context. It works like a "USB-C for AI": a universal standard that allows any AI model to interact with any compatible data source or tool, without the need for custom integrations for each combination.
Before MCP, each integration between an AI model and a data source required custom development: specific APIs, ad-hoc data transformations, and manual context management. MCP solves this by defining a standard protocol with three main primitives: Resources (data the model can read), Tools (actions the model can execute), and Prompts (reusable templates for interactions).
MCP operates with a client-server architecture where the host (the AI application) connects to MCP servers that expose the capabilities of each data source or tool. This allows an AI model to access a data catalog, execute queries, consult documentation, or interact with governance tools without the developer having to build specific integrations for each combination.
Why it matters
Generative AI is transforming how teams interact with data, but without access to the right context, models produce incorrect or incomplete responses. MCP solves the problem of connecting AI to the real organizational context in a standardized way. Instead of copying data manually into prompts or building fragile integrations, MCP provides structured, secure access.
For data governance, MCP is especially relevant because it allows AI agents to interact with data catalogs, glossaries, and lineage programmatically. An AI agent with MCP access to a catalog can answer questions such as "What tables contain customer data?" or "What is the lineage of this KPI?" by querying the metadata directly, not by guessing.
How it works in practice
MCP works with a three-layer architecture. The Host is the AI application (a chatbot, an agent, an IDE). The MCP Client is the component that manages the connection with servers. MCP Servers expose the capabilities of each integration (one server for the data catalog, another for the database, another for the ticketing system).
When a model needs information, the MCP client queries the available servers, retrieves the relevant resources, and includes them as context in the request to the model. When the model needs to execute an action, it uses the tools exposed by MCP servers. All of this occurs with built-in security controls: authentication, authorization, and action logging.
MCP (Model Context Protocol) in Linedat
Linedat exposes its data catalog, business glossary, and lineage through an MCP server, allowing AI agents and compatible tools to query the organization's metadata in a standardized way. This enables use cases such as AI assistants that answer questions about data by querying the catalog directly, without the need for custom integrations.
Related terms
A Data Catalog is a centralized inventory of all of an organization's data assets, with metadata, descriptions, and lineage.
What is Data Governance?Data Governance is the framework of policies, processes, and roles that ensures data quality, security, and correct use across an organization.
What is Metadata Management?Metadata management organizes data about your data: schemas, descriptions, lineage, and classifications in a centralized system.
What is Data Lineage?Data Lineage is the traceability of data from its origin to its destination, showing all transformations and dependencies along the way.