# VelesDB Memory

An offline memory layer for your agent: it keeps facts, retrieves them by meaning, links them together and can show the evidence behind a recall. No cloud, no API key.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed · ★ 100
- Author: cyberlife-coder
- License: not stated · policy=link-only · description only; the file stays with the author
- Source: https://github.com/cyberlife-coder/VelesDB
- Updated: 2026-10-24
- HTML: https://affpapa.org/en/nevibecode/mcp/velesdb-memory

## What it does

What gets stored are single statements, not whole chats, and they go into a local file store. Retrieval works on meaning rather than wording: a local embedding model maps text to coordinates, so a question about which port you settled on still reaches the record even when no word overlaps. A graph holds the links between facts, typed columns hold the exact values, and one query language covers all of it.

A recall can surface the evidence it was built from. Context compression reports a rule id, a reason and a risk level for each decision. Nothing in the write path calls a model, which keeps results repeatable.

You need a Rust toolchain for `cargo install velesdb-memory`, or the Node package, or the prebuilt `.mcpb` bundle published in the official MCP Registry. Vector dimension comes from your embedding model and similarity semantics set the metric. A collection locks both at creation; changing either means a fresh collection with documents re-indexed.

## Install

### Claude Code (native platform)

**Install**
- The MCP registry has no launch command for it: see the repository README

## When to use it
- carry agent memory across sessions without a hosted service
- retrieve an earlier decision by meaning instead of keyword match
- connect facts in a graph and inspect what a recall is based on
- shrink an oversized tool result before it lands in the transcript
- restore working context after a session restart or compaction

## Example prompts
- Remember that the API listens on port 6333 because 3000 was taken by the web UI
- Which port did we pick for the API?
- Show me the records this answer came from

## Similar
- [REA](https://affpapa.org/en/nevibecode/mcp/rea.md) — MCP server · ★ 386989
- [GitHub](https://affpapa.org/en/nevibecode/mcp/github.md) — MCP server · ★ 33918
- [Netdata](https://affpapa.org/en/nevibecode/mcp/netdata.md) — MCP server · ★ 81115
- [browser-use](https://affpapa.org/en/nevibecode/mcp/browser-use.md) — MCP server · ★ 119160
- [Codebase Memory](https://affpapa.org/en/nevibecode/mcp/codebase-memory.md) — MCP server · ★ 47392
- [mcp-server](https://affpapa.org/en/nevibecode/mcp/mcp-server.md) — MCP server · ★ 43872

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md