🦞 ViBo MCP Server

Memory for AI agents — persistent memory (L1/L2/L3 encryption), web-search savings and thread memory. Works with any MCP client: Claude Desktop, Cursor, OpenClaw, Windsurf, Codex, and more.

Install

npm install -g @vibo-dev/vibo-mcp

Configuration

Add to your MCP client config (Claude Desktop example):

{
  "mcpServers": {
    "vibo": {
      "command": "npx",
      "args": ["-y", "@vibo-dev/vibo-mcp"],
      "env": {
        "VIBO_API_KEY": "YOUR_VIBO_KEY"
      }
    }
  }
}

Get a key: https://wwwvibo.com — free 2-day trial, then $5/month.

Tools

Tool	Description
`memory_search`	Find relevant facts (returns token savings)
`memory_add`	Save a fact (dedup by exact match)
`memory_usage`	Your real savings statistics
`thread_memory`	Thread: add / compress / ask / context

Example

Agent: "What does client Anna prefer?" → memory_search("Anna preferences") → • [L1] client-anna: Anna likes coffee without sugar, order #42 → 💾 Saved 13,452 tokens (97.5%)

Honest numbers

Memory: 97.5% fewer tokens on 118 facts (grows with memory, 50-150× on 10K+).
Web search: 99.6% (measured 47,443 → 186 tokens).
Threads: -72%.
Secrets (L3) never reach the LLM — encrypted by design.

Related projects

CloudArc — pack a 10 GiB folder into one .vibo archive at a flat ~26 MiB peak RSS, and read its index over HTTP without fetching the payload (Apache-2.0). Docs site
memory-shield — poisoning defense for agent memory (MIT).
ViBo-memory — persistent agent memory with L1/L2/L3 encryption.

Integrations

cc-connect (bridge agent CLIs to messengers): docs/cc-connect.md

Vibo