Memory Box

Overview

Memory Box is MemVerge’s consumer-facing personal AI memory application, described in VOL.001｜从模型到记忆，AI竞争的新战场已经出现｜对话 MemVerge CEO Charles as the first commercial product built on Memory Machine.

Current Profile

The product aims to index scattered personal data and let users query it without manually finding and uploading individual files. Its proposed architecture combines an embedded chat agent, local models, selective remote inference, browser integration, and a planned MCP server. This is a product-team description; the source does not independently verify privacy, retrieval, or adoption outcomes.

Key Characteristics

• Targets heavy AI users and some small-team use rather than developers alone.
• Connects local files, cloud storage, applications, email, chat history, and model conversations.
• Supports local and hybrid execution according to sensitivity and task difficulty.
• Aims to become a primary personal AI interface while remaining callable from other tools.
• Plans free and paid subscription tiers plus broader mobile availability.

Evidence

Personal-data interface

• VOL.001｜从模型到记忆，AI竞争的新战场已经出现｜对话 MemVerge CEO Charles says users should be able to ask about work, finance, or life data without knowing which file holds the answer.

Hybrid execution

• VOL.001｜从模型到记忆，AI竞争的新战场已经出现｜对话 MemVerge CEO Charles describes local handling for private or easier tasks and external-model use for harder non-private work or constrained devices.

Interoperability and commercialization

• VOL.001｜从模型到记忆，AI竞争的新战场已经出现｜对话 MemVerge CEO Charles mentions a browser extension, planned MCP server, PC and Mac support, planned mobile releases, and tiered subscriptions.

Qualifications

• Retrieval quality, zero-retention claims, model routing, context limits, subscription details, and future releases are unverified product claims or plans.
• Centralizing personal information can increase the consequence of incorrect access control, stale memory, or compromised credentials.

What Changed

• Created a product profile centered on personal-data access, hybrid inference, and interoperability.

Relationships

• MemVerge - developer company.
• Charles Fan - CEO and product spokesperson.
• Memory Machine - underlying memory layer.
• Local-First Memory Layer - architectural pattern the product claims to implement.
• AI Memory Lifecycle - retrieval and long-term maintenance problem the product must address.