MCP as a Standardized Cognition Identity Interface for Self Hosted Identity Kernels

This paper proposes using Anthropic's Model Context Protocol (MCP) as a standardized interface between artificial agent identity and cognition engines, extending the Self-Hosted Identity Kernel (SHIK) framework (Kingsley, 2025). By exposing the identity kernel as an MCP server, any MCP-compatible language model can read from and write to an agent's persistent self through tool calls, achieving model independence as a protocol-level guarantee rather than an architectural aspiration. The paper maps SHIK identity tuple components to MCP tool specifications, describes injection and extraction workflows as MCP client middleware, and discusses implications for portable agent identity, inter-agent handshakes, edge deployment, and governance in heterogeneous multi-model environments.