In development

Thomas Memory — Personal Intelligence MCP

Cross-model memory, reusable skills and secure MCP interfaces.

Overview

A personal intelligence platform that gives different AI models a shared, governed memory: reusable skills, prompts and retrieval exposed through secure Model Context Protocol (MCP) interfaces.

Problem

Knowledge and working context get trapped inside individual AI tools. Moving between models means re-explaining context, and there is rarely a clear boundary between private memory and anything published.

Architecture

AI clients connect through an authenticated MCP interface. A memory service handles retrieval over stored notes, skills and prompts, and an approval workflow governs what can ever move from private memory to public content. This portfolio is deliberately separate: it never reads private memory.

Layers
AI CLIENTSSECURE INTERFACEMEMORY COREGOVERNANCEMCPproposals onlyMultiple AI modelsAuthenticated MCPinterfaceRetrievalSkills & promptslibraryPrivate memory storeHuman approvalworkflow

AI models connect through an authenticated MCP interface, which routes to retrieval and a skills and prompts library. Retrieval reads the private memory store. Anything leaving private memory must pass a human approval workflow.

Components — select to trace connections

Select a component to see what it does and how it connects.

Challenges

  • Strong separation between private memory and public content
  • Authenticated, least-privilege MCP access across clients
  • Making skills and prompts reusable across different models

In development. Only the public, high-level architecture is shown.