Model Context Protocol (MCP)
The Model Context Protocol (MCP) is an open protocol that standardizes how AI applications connect to external tools and data sources. Introduced by Anthropic in November 2024 and now developed under open community governance, it defines JSON-RPC 2.0 communication between hosts (LLM applications), clients (connectors inside the host), and servers (services exposing tools, resources, and prompts), so one integration works across any MCP-capable agent.
Before a shared protocol, every AI application that wanted to reach every external system needed a bespoke integration — an N-by-M matrix of connectors. MCP collapses that: a service ships one MCP server, and every MCP-capable host can use it. The design borrows from the Language Server Protocol, which did the same standardization for editors and programming languages.
For an API producer, MCP is the clearest bet on how finished agent integrations will call APIs at runtime, and shipping an MCP surface is a concrete discovery signal. It leaves the documentation problem in place, though: the coding agents that scaffold integrations still have to understand your endpoints, auth, and request shapes from public docs before any tool surface enters the picture, and the MCP server itself is a public surface an agent must find and understand. The discovery problem moves into the manifest and the registry rather than disappearing.
MCP presence counts toward Discry's discovery dimension through two checks: a fetch of the well-known MCP manifest at /.well-known/mcp.json, and a probe of public MCP registry listings for the API's name. Both results publish per profile with their details. MCP surfaces are not yet behaviorally measured; the methodology states that as a known limitation and the direction of the next edition.
What does an AI agent make of your API?
Find out in about a minute — no signup.
Discry your API — free