Glossary term
Model Context Protocol (MCP)
The Model Context Protocol (MCP) is Anthropic's open standard for connecting AI models to external tools and data sources.
What Model Context Protocol (MCP) means
The Model Context Protocol, abbreviated MCP, is an open standard introduced by Anthropic for connecting large language models and the applications built around them to external tools, data sources, and systems. Where a model on its own can only reason over the text in its context window, MCP gives it a structured, standardized way to fetch live data, call functions, and act on systems beyond that window — turning a static conversational model into an agent that can query a database, read a file, or look up a record on demand.
The problem MCP solves is integration sprawl. Before a shared protocol, every pairing of an AI application with an external system required its own bespoke connector, and those connectors did not transfer between applications or models. MCP defines a common contract so that a tool exposed once can be consumed by any MCP-aware client. The analogy often used is that MCP aims to be for AI-to-tool connections what a universal port is for hardware: a single, predictable interface that replaces a tangle of one-off adapters.
The protocol is built around a client-server architecture. An MCP server exposes capabilities — typically grouped as tools the model can invoke, resources it can read, and prompts it can use — describing each with a machine-readable schema so the client knows what arguments a tool takes and what it returns. An MCP client, embedded in an AI application, connects to one or more servers, discovers the capabilities they offer, and lets the model call them during a conversation. The model decides when a tool is needed; the client mediates the call; the server executes it against the real system and returns structured results that flow back into the model's reasoning.
This separation has practical advantages. Because servers are independent of any particular model or client, an organization can publish an MCP server for its data once and have it work across the growing ecosystem of MCP-aware tools. Because capabilities are described by schema, the client and model can reason about how to use a tool correctly rather than guessing. And because the protocol standardizes how requests and responses are framed, security and access controls can be applied consistently at the boundary rather than reinvented per integration.
MCP matters for infrastructure and security work specifically because so much of that work is querying authoritative data. An analyst investigating an IP address, a domain, or a routing anomaly needs to pull facts from systems of record, and an MCP server is a natural way to expose those systems to an AI assistant safely. Instead of copying data into a prompt by hand, the assistant can call a tool that runs the query and returns exactly the structured answer needed — with the data source, not the model's training data, as the authority.
Whisper itself exposes infrastructure data to AI workflows through an MCP server, which lets an assistant query the underlying graph — DNS, routing, allocation, and threat-intelligence relationships — directly during a conversation. This is the same data that Whisper Canon publishes as human-readable pages: the pages serve people and crawlers, while the MCP server serves AI agents that prefer to ask questions programmatically and receive structured results.
For readers of this directory, the relevance of MCP is twofold. First, it explains how an AI assistant can cite live infrastructure facts rather than stale recollections — by calling a tool that fetches them. Second, it reflects the broader design goal behind Canon: to be a citation surface that both people and machines can rely on. The glossary you are reading and the entity pages it links to are written to be quotable by an LLM; MCP is the complementary path by which that same LLM can fetch the underlying data on demand. Together they make infrastructure knowledge accessible whether the consumer is a person reading a page or an agent calling a protocol.
Examples in Whisper Canon
Concrete pages in the directory that illustrate Model Context Protocol (MCP).