Skip to content

[Proposal] Memory for MemOS agents — 97.5% fewer tokens (ViBo) #2244

Description

@vnbochkarev-netizen

Hi MemOS team,

Love MemTensor/MemOS (10706⭐) — impressive work. I noticed agents built on it
often hit the same wall: memory between sessions is missing or too
expensive to include fully on every request.

ViBo — memory for AI agents, one skill, three capabilities:

  • 🧠 Encrypted L1/L2/L3 memory: secrets NEVER reach the LLM
  • 🌐 Web search savings: 47,443 → 186 tokens (99.6% measured)
  • 💬 Thread memory: conversations -72%, details restored on demand
  • Measured: 118 facts → 263 tokens instead of 13,775 = 97.5% fewer

Integrates via Python API or MCP — works with any agent/framework
in minutes. Single .web file, fully portable, no vendor lock-in.

Model: $5/month · 2-day free trial (key built-in).
Site: https://wwwvibo.com · Docs: https://github.com/vnbochkarev-netizen/ViBo-memory

Open to integration, co-marketing, or referral partnership (30% recurring).
Honest limits included: code gen doesn't save — we say so publicly.

Best,
ViBo team

Metadata

Metadata

Labels

area:coreMOS 编排层 / 框架底座 / 跨模块问题status:needs-triageNeeds initial triage | 需要初步判断 & 问题复现types:enhancementNew feature or improvement | 新功能或改进

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions