MCP (Model Context Protocol)
An open protocol for connecting AI models to tools and data sources through a standard interface. Instead of custom integrations per model per tool, MCP servers expose capabilities any compliant client can use.
The integration matrix it collapses
Before a standard existed, connecting M models to N tools meant building M multiplied by N bespoke integrations, each one maintained by somebody. MCP turns that into M plus N: tool authors expose one server, model clients speak one protocol, and any compliant pair works without either side knowing about the other.
Servers, clients and the contract between them
An MCP server advertises tools, resources and prompts with explicit schemas. A client — an agent runtime, an IDE, a chat application — discovers what is available and invokes it over a standard transport. Because the contract is declared rather than assumed, a client can reason about what it is calling before it calls it.
What it changes for agent builders
Your runtime stops needing to ship every integration itself, and capability becomes something you install rather than something you wait for. It also gives operators a single place to reason about permissions, which matters more as the number of connected tools grows.
In AGNT
AGNT speaks MCP natively — plug MCP servers in as tool sources alongside built-in nodes.
MCP (Model Context Protocol) — common questions
Who created MCP and is it open?
Anthropic released the Model Context Protocol as an open standard in 2024. It has since been adopted well beyond its origin and functions as a cross-vendor standard.
Does AGNT support MCP?
Natively. MCP servers connect as tool sources and their tools appear in agent toolkits alongside built-in nodes, with the same permissions and receipts.
Is MCP a replacement for APIs?
No — it sits above them. An MCP server usually wraps existing APIs and presents them in a form models can discover and call reliably.