Skip to main content

Zero-knowledge encrypted memory library for personal AI agents

Project description

VaultMem

A zero-knowledge encrypted memory library for personal AI agents.

What This Is

VaultMem is an embeddable open-source library that any platform developer can integrate to give their users persistent, encrypted, portable AI memory — where the platform operator cryptographically cannot read user data.

This is the concrete implementation of Layer 1 (Personal Storage Layer) of the Memory-as-Asset framework proposed by Pan, Huang & Yang (arXiv:2603.14212, March 2026).

The Problem

Every existing AI memory library (mem0, Zep, LangMem, Letta) stores user memories in plaintext on the platform's servers. The platform owns your memory. Switch platforms and you start from zero.

VaultMem flips this:

  • User controls the encryption key (passphrase → Argon2id → session key)
  • Key lives only in RAM during active sessions — never written to disk
  • Platform developer embeds the library but cryptographically cannot read memories
  • Users export their encrypted vault and take it anywhere

Repository Structure

vaultmem/
├── research/           # All research gathered during ideation
│   ├── 01_existing_memory_systems.md
│   ├── 02_memory_as_asset_paper.md
│   ├── 03_privacy_taxonomy.md
│   ├── 04_voice_interfaces.md
│   ├── 05_author_contacts.md
│   └── 06_related_papers.md
├── paper/              # The arXiv paper being written
│   └── outline.md
├── design/             # Architecture and design decisions
│   └── vaultmem_architecture.md
└── README.md

Status

  • Research phase complete
  • Architecture designed
  • Paper outline drafted
  • Paper sections in progress
  • Python prototype
  • arXiv submission
  • Outreach to MAS paper authors

Related Work

  • Memory as Asset (arXiv:2603.14212) — the framework paper this implements
  • Engram (github.com/EvolvingLMMs-Lab/engram) — closest existing project, MCP tool not library
  • OpenSecret (opensecret.cloud) — TEE-based alternative approach

Paper Target

arXiv preprint → ICML 2026 workshop or NeurIPS 2026 workshop

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

vaultmem-0.1.1.tar.gz (29.3 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

vaultmem-0.1.1-py3-none-any.whl (27.6 kB view details)

Uploaded Python 3

File details

Details for the file vaultmem-0.1.1.tar.gz.

File metadata

  • Download URL: vaultmem-0.1.1.tar.gz
  • Upload date:
  • Size: 29.3 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.1

File hashes

Hashes for vaultmem-0.1.1.tar.gz
Algorithm Hash digest
SHA256 083b030cb33ac582cebf365533bebb3124d6a4299e867cd828758664a528262a
MD5 6cafcf394395003ef94f60ccbebfa289
BLAKE2b-256 737611e57f6a27e7e65eef780ae24e73bc5625e56e849ec036c36223afb842cc

See more details on using hashes here.

File details

Details for the file vaultmem-0.1.1-py3-none-any.whl.

File metadata

  • Download URL: vaultmem-0.1.1-py3-none-any.whl
  • Upload date:
  • Size: 27.6 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.1

File hashes

Hashes for vaultmem-0.1.1-py3-none-any.whl
Algorithm Hash digest
SHA256 63ef078fc2130e1c51f7879e67c491755b7f0a833715d061d7057f4aee59fb52
MD5 828f627f067ff4abd48c1d5f6cd7320f
BLAKE2b-256 368b1d9794c62254c4f5e28f9f2931d7d5d130bdf5a67d6a405db0d93ec2ae20

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page