CL Compound Leverage Platform
AI Team Plans Talk with a Placement Specialist
Platform/Model Context Protocol and AI Teams

Model Context Protocol and AI Teams

Model Context Protocol (MCP) can provide a standardized mechanism for connecting AI execution environments to approved data, tools, and computational capabilities.

MCP is useful within the AI Team architecture, but the architecture does not depend on MCP.

MCP is an implementation mechanism

The durable architectural concepts are the job, skill, workflow, capability request, authorization boundary, and customer context.

MCP can implement parts of those relationships today. Another protocol or model-native mechanism can replace it later without requiring the AI Team’s business architecture to be redesigned.

Customer MCP

A customer-controlled MCP implementation can be one component of the Customer AI Control Plane.

It may expose:

  • approved enterprise data
  • internal search or retrieval
  • business applications
  • customer tools
  • organizational routing information
  • governed actions

The customer can determine what is exposed and under which controls.

Customer MCP is optional. A deployment can use native connectors, APIs, files, model-platform capabilities, or other mechanisms instead.

Compound Leverage MCP

Compound Leverage can also use MCP to expose controlled capabilities without distributing proprietary implementation logic into a plugin or customer environment.

Examples can include protected computations, evaluation functions, scoring services, or other capabilities where the Digital Employee needs the result but does not need the underlying implementation in its context.

Two different boundaries

It is useful to keep these concepts separate:

  • Customer-controlled services provide governed access to customer resources.
  • Compound Leverage-controlled services can expose protected platform capabilities.
  • Digital Employees and skills request the capability they need without requiring the implementation to live in the persona or prompt.

Why this matters for model evolution

As AI runtimes become more capable, they may absorb capabilities that currently require separate tools or orchestration calls.

The AI Team architecture can take advantage of that change because MCP is treated as an execution option rather than the foundation of the workforce model.

See AI Agent Tools and Enterprise Integrations for the broader capability layer.