Get started

MCP concepts

Model Context Protocol is an open standard from Anthropic for connecting AI assistants to external tools, data, and APIs. Three primitives, one transport.

The three primitives

Tools — functions the AI can invoke. Each has a name, a description, an input schema (JSON Schema or Zod), and a server-side implementation. The AI decides when to call them based on the description.
Resources — read-only data the AI can fetch. URI-addressed. Think files, documents, database records.
Prompts — parameterized prompt templates the server exposes to the client. Used for orchestration patterns the server author wants to encourage.

Transport: Streamable HTTP

Every generated server speaks Streamable HTTP (MCP spec 2025-11-25). The previous SSE transport was deprecated in June 2025 and is not supported here. One HTTP endpoint at /mcp, optionally negotiating a long-lived stream viatext/event-stream when the server wants to push updates.

single request

POST /mcp HTTP/1.1
Authorization: Bearer <jwt>
Content-Type: application/json
Accept: application/json, text/event-stream

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/list"
}

Session lifecycle

initialize — client sends protocol version + capabilities, server returns its info and assigns a session id via the mcp-session-id header.
notifications/initialized — client confirms readiness. Server is now free to push notifications.
tools/list, tools/call, resources/list,prompts/list — the work.

Why MCP and not just REST

REST APIs need bespoke OpenAPI integration per client. MCP standardizes the discovery, invocation, auth and streaming so any spec-compliant client picks up any spec-compliant server with zero glue code. That's the entire point.