The Model Context Protocol is more than a buzzword — it is a wire format and capability contract between AI clients and external systems. Understanding the three layers (transport, tools, resources) helps you evaluate hosted providers like MCP360 versus building your own integration stack.
Transport Layer
MCP supports two primary transports:
Both transports carry the same JSON-RPC-style messages. Switching transports does not change your tool definitions.
Tools: Structured Actions
Tools are the write and read operations your AI can invoke. Each tool declares:
When a user asks "pause my worst-performing Meta campaign," the model selects update_campaign_status, fills in the campaign ID and desired state, and the MCP server executes against the real API.
Resources: Readable Context
Resources expose read-only context — files, database rows, configuration snapshots — that the model can fetch without side effects. Tools perform actions; resources provide grounding data.
Discovery at Connect Time
On connection, the client calls tools/list and resources/list to learn what the server offers. This is why one MCP360 API key unlocks dozens of ad management tools without per-prompt configuration.
Why Businesses Use Hosted MCP
Building OAuth flows, token refresh, rate limiting, and API versioning for Meta, Google, and Shopify individually takes months. Ad Platform MCP and other MCP360 services ship maintained tool catalogs so your team configures an endpoint, not an integration project.
Start with What is MCP? for the business case, then connect your first client via Claude Desktop setup.