MCP

MCP — Model Context Protocol — is an open protocol for exposing tools, resources, and contextual capabilities to an AI application. An eqlx system may use MCP-compatible servers where they fit the customer's stack, authorization model, and operating boundary. MCP is optional; an agentic system does not require it merely because it contains an agent.

Protocol is not policy

MCP defines how clients and servers exchange capabilities; it does not decide which business action should be allowed. Stable validation, authorization, rate-limit handling, and domain rules should live in deterministic application or server logic. Model judgment is reserved for decisions that genuinely require it and is still bounded by those controls.

How the interface is selected

The required capability and operating boundary determine whether an existing MCP server, a direct API integration, or another connector shape is appropriate. Any custom connector included in scope is documented with its permissions, schemas, error behavior, and handoff terms.

What the customer reads

An MCP server describes the inputs and outputs of the capabilities it exposes. The client-server contract, supported protocol revision, authorization boundary, and error behavior should be versioned and tested. A vendor API change may still require server and integration updates; MCP reduces coupling but does not remove it.